Lenny's Podcast

The most rational take on AI you’ll hear this year

2559 summary words 11 min summary Watch video

Start with the signal

11 min read

Summary

At-a-Glance

  • Verdict: Watch fully
  • Core thesis: AI is a platform shift on the scale of the internet or mobile, but its eventual products, winners, labor effects, and value distribution remain radically uncertain and will emerge through years of uneven deployment.
  • Why it matters: The discussion provides a strong strategic framework for AI products and agent systems: distinguish tasks from jobs, deployment from raw capability, infrastructure from application value, and temporary pricing disequilibrium from sustainable market power.
  • Best use: Use it to pressure-test AI investment theses, product positioning, model-layer dependence, enterprise deployment plans, and claims about imminent job replacement.

Executive Summary

Benedict Evans rejects both AI dismissal and deterministic superintelligence narratives. His comparison is to the internet in 1997: the underlying shift is unquestionably enormous, software development has already been transformed, but most products do not work reliably enough, adoption outside technology remains shallow, and the eventual winners may not resemble today's leaders. Even if models stopped improving tomorrow, he argues, current capabilities would still take roughly a decade to diffuse through the economy.

Enterprise change will be constrained less by access to models than by organizational deployment. Companies must identify useful workflows, connect horizontal and vertical systems, manage internal politics, redesign processes, and train employees. That makes forward-deployed engineers, consultants, and implementation partners more important in the near term, not obsolete. The right unit of analysis is also the job rather than the visibly automatable task: generating code or slides may become cheap while customer discovery, judgment, accountability, and organizational change remain scarce.

Evans is skeptical that foundation-model labs will capture most of the long-term value. Models currently show little evidence of network effects, several providers may remain competitive, and customers may ultimately treat inference like cloud or telecom infrastructure. If differentiated applications and integrated workflows become the real user experience, value and pricing power should migrate up the stack toward products with distribution, proprietary context, workflow ownership, and brand.

On employment, Evans expects genuine dislocation but rejects forecasts that translate task exposure directly into job losses. Previous technology waves eliminated occupations while lowering costs, expanding demand, and creating work that could not have been predicted in advance. His practical advice is not complacency: people and companies should immerse themselves in the technology, learn where it works and fails, and become unusually capable at applying it rather than resisting it on principle.

Key Takeaways

  • Claim: AI should be treated as a foundational platform shift comparable to the internet or mobile, while presuming radical uncertainty about products, winners, and end states. | Evidence: Evans compares the market to 1997, when internet potential was obvious but debates about whether Excite or Yahoo would win missed the companies and products that ultimately mattered. He notes that today's adoption ranges from technologists replacing Google with local setups to many mainstream users opening an AI tool only every week or two. | Implication: Ken should build strategies that benefit from broad AI adoption while remaining modular on models, interfaces, and channels rather than hard-coding a single forecast about the winning lab or UX. | Caveat: The historical analogy establishes scale and uncertainty, not a precise adoption timeline or proof that AI will follow the internet's market structure.
  • Claim: The primary bottleneck in enterprise AI is deployment and organizational redesign, not access to model intelligence. | Evidence: Evans says a company-wide workflow review requires five to ten people working for one or two months, followed by another project to connect vertical systems with horizontal systems, construct workflows, and train employees. He frames forward-deployed engineers as the modern equivalent of outsourced Accenture developers and explains why OpenAI, Anthropic, private-equity firms, and consultancies are investing in implementation capacity. | Implication: OpenClaw and related agent systems need a deployment layer covering discovery, integration, permissions, training, monitoring, and change management; model access alone is not an enterprise solution. | Caveat: Professional services may themselves become more productive and change shape; current demand for implementation does not guarantee existing consulting economics indefinitely.
  • Claim: Automating a task does not predict whether a job disappears because the task being automated may not be the scarce or valuable part of the job. | Evidence: Claude can generate code, but it cannot by itself determine which customer to serve, which product to build, or how to take it to market. Likewise, a weak AI-generated 75-slide deck is not what clients actually buy from Bain or McKinsey; they buy investigation, customer interviews, political navigation, diagnosis, and organizational action. Evans also notes that accounting employment continued rising through adding machines, mainframes, ERP, PCs, and cloud spreadsheets. | Implication: Workflow design and investment analysis should identify the remaining bottleneck after generation becomes cheap—judgment, context, trust, coordination, distribution, or accountability—rather than estimating a simplistic percentage of work automated. | Caveat: Some occupations really are reducible to a task, as with elevator attendants operating manual elevators, and those jobs can be eliminated directly.
  • Claim: AI will cause labor-market pain, but there is insufficient evidence for an immediate economy-wide job apocalypse. | Evidence: Evans points to repeated historical cycles since 1800 in which technology removed visible jobs but created initially unimaginable ones. He also emphasizes that large-enterprise software sales cycles can take around 18 months and that replacing estates such as SAP can take three, five, or ten years. Current studies show weaker employment among 18- to 24-year-olds, but similar patterns appear across degree status and both AI-exposed and less-exposed fields. | Implication: Ken should expect uneven, sector-specific transitions rather than a single automation event, while watching entry-level hiring and apprenticeship structures as leading indicators of organizational change. | Caveat: Aggregate prosperity does not protect specific workers, towns, or career ladders. Entry-level professional-services roles and the associate-to-partner pyramid may face serious disruption even if total employment ultimately recovers.
  • Claim: Foundation models may become competitive commodity infrastructure, leaving more durable value in applications and workflows above them. | Evidence: Evans argues that models do not currently display obvious network effects and may remain a field of three to ten competing providers. He compares them with mobile telecoms: the global mobile industry earns roughly $1 trillion annually, spends about $200 billion per year on capex, and carries 1,500 to 2,000 times the data volume of 2010, yet telecom stocks have delivered little growth because the differentiated value sits above the network. | Implication: Ken should favor architectures with model routing and substitution, and evaluate businesses on workflow control, proprietary context, distribution, and measurable outcomes rather than privileged access to one foundation model. | Caveat: Evans explicitly calls this a provisional thesis and notes that a 1997 observer would likely have missed Apple's and Google's eventual roles. Future model differentiation, integration, or scale advantages could still create pricing power.
  • Claim: Distribution and defaults become stronger advantages when the underlying AI product is adequate but difficult for mainstream users to distinguish. | Evidence: Google can push Gemini through its existing surfaces, while Meta achieved substantial reported usage by placing its assistant across its products even when technologists had dismissed the underlying model. Evans compares this with browsers, where the thin product converged around a basic input-output interface and Microsoft used distribution to gain share, although browser dominance did not ultimately capture the value further up the web stack. | Implication: Agent products need embedded distribution and workflow-level retention, not merely a better general chatbot. The defensible surface is likely the place where work happens and context accumulates. | Caveat: Distribution can win usage without capturing economics; Microsoft's browser victory did not translate into ownership of web applications.
  • Claim: The largest opportunities will come from redefining the activity around AI rather than performing the old activity faster or more cheaply. | Evidence: Recorded-music revenue fell by about half between 2000 and 2015 as individual tracks escaped the $15 CD bundle, then recovered to roughly 75% of its inflation-adjusted peak as $15 monthly subscriptions offered access to essentially all music. Evans uses Spotify, Uber, and Airbnb to distinguish digitizing an old workflow from creating a new product category. | Implication: Ken should test whether an agent concept merely accelerates an existing task or enables a new operating model, service level, market, or pricing structure that was previously uneconomic. | Caveat: Category disruption remains highly specific: Uber substantially changed taxi markets, while Airbnb created a major adjacent category but had only a marginal effect on hotels, especially business travel.

Detailed Brief

AGI language is too unstable to anchor operating decisions

  • Claims: Evans argues that neither human intelligence nor the mechanisms behind current model performance are understood well enough to forecast human-level intelligence confidently.; Definitions of AGI are being revised from machine consciousness or general human equivalence toward completing some percentage of economically valuable work, which makes apparently precise AGI claims difficult to compare.; Transformative economic impact does not require AGI: today's models could stop improving and still support a decade of deployment and product creation.
  • Evidence: He cites Larry Tesler's formulation that AI is whatever machines cannot do yet; once a capability works, people relabel it as ordinary software.; An IBM mainframe in 1975 already performed a meaningful share of economically valuable work previously done by people without satisfying stronger notions of general intelligence.; Evans describes most AGI timelines as 'vibes forecasting' because there is no accepted theory indicating how much better current systems will become.
  • Caveats: The absence of a reliable theory does not establish that AGI is impossible; it means high-confidence timing and labor forecasts are not justified by the evidence presented.
  • Implications: Scenario planning should separate capability milestones that can be measured from semantic claims about AGI.; A product or investment thesis should work under current-model deployment and not depend entirely on speculative superintelligence.

Backlash combines legitimate local harms with weak generalizations

  • Claims: Anti-AI sentiment is not one issue but a mixture of energy costs, employment anxiety, creator displacement, low-quality generated content, privacy concerns, and broader distrust of technology companies.; AI can amplify abuse by reducing the skill, time, and cost required to create harmful content, even when a primitive version of the capability existed before.; Technical systems can ruin lives through institutional overconfidence and faulty automation without requiring advanced AI.
  • Evidence: A Livermore Lab estimate cited by Evans put U.S. data-center water consumption at approximately 0.017% of total U.S. water consumption. He separately estimates data centers at around 5% of U.S. energy use, potentially increasing by one percentage point annually for five years.; He uses deepfake sexual imagery as an example: Photoshop existed previously, but a teenager can now generate explicit images or video targeting many classmates in a single afternoon.; In the UK Post Office scandal, bugs in a Fujitsu point-of-sale system produced false cash shortfalls; hundreds of franchise operators were prosecuted or imprisoned, with bankruptcies, lost homes, and suicides, while officials denied the software faults.
  • Caveats: National water-use percentages can conceal severe local constraints when a data center competes for a small community's limited water supply.; Some labor and environmental claims remain unsettled because model labs disclose little meaningful usage data and researchers must infer effects from surveys and government statistics.
  • Implications: Public trust will depend on local impact, recourse, auditability, and institutional behavior—not only model accuracy.; AI governance should account for scaled misuse and automation bias rather than dismissing concerns because analogous tools existed previously.

The blank chatbot is an incomplete product interface

  • Claims: A general chatbot confronts users with both a blank screen and a jagged capability frontier: they may not know what to ask, whether a task will work, or whether the result is correct.; Useful AI will often disappear into specific product features and eventually be perceived as ordinary automation rather than as an AI destination.; Breakout consumer AI products remain constrained partly by inference cost because companies cannot necessarily offer a free product to 50 million users and monetize later.
  • Evidence: Evans uses AI for proofreading, image-based apartment redesign, and voice transcription, but not for the precise information-retrieval work central to his analyst role because hallucinations remain problematic.; He dictates much of his writing into Apple Notes and treats the automatic transcription as a feature rather than a separate AI product.; Apple's 2024 vision combined tool use, on-device processing, app intents, and a personal assistant, but Evans notes that delivering it without prompt injection, hallucinations, or broken integrations across thousands of apps proved difficult.
  • Caveats: Evans's own use pattern reflects an unusual synthesis-heavy analyst job and should not be generalized to all professions.; Application wrappers still inherit model reliability, security, and marginal-cost constraints.
  • Implications: The strongest agent UX may be a constrained workflow with explicit tools, context, validation, and escalation rather than an open-ended conversation.; Successful AI features may become invisible, so product measurement should focus on completed outcomes rather than chatbot engagement alone.

Notable Concepts & Terms

  • Presume radical uncertainty: Treat AI's scale as clear but its market structure, winners, products, timelines, and labor effects as unresolved.
  • Jagged frontier: AI works extremely well on some adjacent tasks and fails unpredictably on others, making capability discovery and result verification part of product design.
  • Task versus job: The visible output being automated may be only one component of the value customers or employers actually purchase.
  • Jevons paradox: Lowering the cost of an activity can expand demand and total usage instead of proportionally reducing spending or employment.
  • Lump of labor fallacy: The economy does not contain a fixed quantity of work; automation can create new demand, industries, and occupations.
  • Forward-deployed engineer: An implementation role that embeds technical talent with customers to discover workflows, integrate systems, and make AI operational.
  • Commodity infrastructure: Evans's provisional model for foundation models: technologically sophisticated and capital-intensive, but potentially competitive, substitutable, and priced near marginal cost.
  • Old thing, but more: The first stage of a platform shift replicates existing activities; the larger opportunity arrives when products redefine the activity itself.

Operator Notes / Why Ken Should Care

  • Create a model-portability benchmark for OpenClaw covering quality, latency, token cost, tool-use reliability, and migration effort across at least three providers.
  • Track anomalous inference economics separately from steady-state unit economics; the cited OpenClaw operator spending $1.5 million on tokens in one month is a warning against treating current consumption patterns as durable.
  • Package enterprise deployments as a defined discovery-to-production engagement with workflow mapping, system integration, permissions, training, evaluation, and escalation ownership.
  • Instrument task completion, correction rates, human escalation, active use, and realized labor-time savings internally rather than relying on sparse public adoption statistics.
  • Audit each planned agent workflow for the true scarce input after generation becomes cheap, and avoid automating outputs whose value depends primarily on unencoded judgment or organizational authority.
  • Require provenance, audit logs, reversible actions, and human recourse for workflows that can affect employment, money, legal status, access, or reputation.
  • Prioritize embedded entry points and channel partnerships where users already work; avoid making a generic chatbot the sole acquisition and retention surface.

Source/Metadata

  • Title: The most rational take on AI you’ll hear this year
  • Transcript words: 23099
  • Duration seconds: 4790
  • Timestamp note: No timestamps or chapter markers were present. The supplied transcript contains substantial duplicated passages, speaker-label artifacts, advertisements, and repeated outro material.
Search this video Resume transcript at 0%
Full transcript 15868 words · 102 min read
0:00

SPEAKER_01

My most controversial opinion is that I think AI is as big a deal as the internet or mobile, and only as big a deal as the internet or mobile. What's your guess on the coming jobpocalypse? Every time we have a new technology, it automates away a bunch of jobs, and then that automation unlocks a bunch of new jobs. And you don't know the new job because it doesn't exist yet. We've had that process over and over again.

0:17

SPEAKER_00

Even just looking at the most advanced AI companies—Big, OpenAI— everyone's increasing headcount.

0:22

SPEAKER_01

You talk to these doomers on Twitter, and they act as though every big company is going to buy ChatGPT tomorrow, and then, in two weeks' time, they'll fire all their staff. These people are morons. You can't predict which things are going to be exposed. You can't look at a senior partner at a law firm and say, “Well, 17% of their work could be automated.” This is horseshit. I'm curious if you're following the anti-AI sentiment. It's a big fuzzy mess. Yes, this will change a bunch of stuff, and we'll need to worry about it. But that's a constant. We've always had that.

0:50

SPEAKER_00

What would be a couple of things you recommend people do to be more successful in this future?

0:56

SPEAKER_01

Don't stick your head in the sand and say, “I hate all of this stuff.” That gives you a great feeling of moral superiority, and you can go on Bluesky and shout at everybody about how evil AI is. Great, I'm happy for you. But that's not going to help. What helps is you diving into this and coming out understanding what you can do with it.

1:14

SPEAKER_00

Today, my guest is Benedict Evans. Benedict was a longtime partner at a16z as their in-house analyst and resident thinker. Before that, he was a longtime equity researcher. And for the past six years, he's been an independent analyst tracking the most important tech trends and sharing what he's learning. Most recently, as you'd expect, he's spending all his time on how AI is changing our lives. And in his words, AI is eating the world. In this conversation, we go deep on what we're still not pricing in about the impact that AI is going to have on our lives and our work, the rise of anti-AI sentiment, the impact on jobs,

1:51

SPEAKER_00

where in the value chain most of the value will accrue, and tons more. If you are worried about AI or just confused about where things are heading, this conversation will teach you a lot and also make you feel better. Before we get into it, don't forget to check out LennysProductPass.com for a year free of some of the most amazing, hottest, most well-crafted AI products in the world, available exclusively to Lenny's Newsletter subscribers. With that, I bring you Benedict Evans.

2:23

SPEAKER_00

Benedict, thank you so much for being here. Welcome to the podcast.

2:27

SPEAKER_01

Thank you for inviting me.

2:28

SPEAKER_00

You just put out this deck called AI Is Eating the World. I want to ask you the flip side of this. We all know it's a big deal. Knowing that, what do you think people are still not fully pricing in when they think about the change that they're going to experience in their lives and their work?

2:46

SPEAKER_01

An interesting way of thinking about it: I did a podcast last year with someone where I said, my most controversial opinion is that I think AI is as big a deal as the internet or mobile, and only as big a deal as the internet or mobile, because clearly there's a bunch of people in tech who think this is more like the Industrial Revolution or something. And there are a whole bunch of people underneath saying, “Well, he thinks this is just as big as—does he not understand how big this is?” Smartphones were quite a big deal. The internet was quite a big deal. We wouldn't be doing this if it wasn't for the internet. So there's one layer of—but then if you dig into that,

3:21

SPEAKER_01

if you're going to make the internet comparison, it's like we're in 1997. It's very exciting. Most stuff doesn't work yet. Most of the stuff that people are going to do hasn't been built yet. And it's not really clear how much of it is going to work when it does work. And the people who have already got it, who have already taken whichever pill it is— I forget which— imagine that everybody in the world is already there. And the truth is you've got this very wide distribution. So there are people in tech who bought their cluster of Mac minis and don't use Google anymore. And then you look outside tech, setting aside the idiots who think that this isn't real.

4:05

SPEAKER_01

Most people who are using this are using this every week or two, maybe. So you've got that spread of adoption and that spread of maturity in how well this works. And then within that, you can make specific points about: Well, how are the models going to work? And do the model labs have pricing power? And where's the value going to be? And has OpenAI won the whole thing? Or has Anthropic got it this week? And so then you can get into calling those races where, again, it's like being in 1997 and saying, “Well, is it going to be Excite or Yahoo?” And the answer was no, generally. So there's a fractal point here. There's the super-high level

4:49

SPEAKER_01

that this is going to change absolutely everything. I don't think it's particularly productive to say, “Well, is it 20% bigger than the internet or 100%?” Those aren't productive conversations. But it's one of those fundamental changes. But then you don't know how any of it's going to work. In fact, I just published this. I do a presentation every six months, and I just published one yesterday. And one of the comments was, “Benedict, this is 80 slides of saying we don't know,” which is slightly facetious, but also true.

5:15

SPEAKER_00

This episode is brought to you by our season's presenting sponsor, WorkOS. [SPEAKER_01] Well, is it 20% bigger than the internet or 100%? [SPEAKER_01] Those aren't productive conversations. [SPEAKER_01] But then you don't know how any of it's going to work. [SPEAKER_01] In fact, I just published this. [SPEAKER_01] I do a presentation every six months, [SPEAKER_01] and I just published one yesterday. [SPEAKER_01] And one of the comments was, [SPEAKER_01] “Benedict, this is 80 slides of saying we don't know,” [SPEAKER_01] which is slightly facetious, [SPEAKER_01] but also true. This episode is brought to you by our season's presenting sponsor, WorkOS.

5:53

SPEAKER_00

[SPEAKER_01] Well, is it 20% bigger than the internet or 100%? [SPEAKER_01] Those aren't productive conversations. [SPEAKER_01] But it's one of those fundamental changes. [SPEAKER_01] But then you don't know how any of it's going to work. [SPEAKER_01] In fact, I just published this. [SPEAKER_01] I do a presentation every six months, [SPEAKER_01] and I just published one yesterday. [SPEAKER_01] And one of the comments was, [SPEAKER_01] “Benedict, this is 80 slides of saying we don't know,” [SPEAKER_01] which is slightly facetious, [SPEAKER_01] but also true. This episode is brought to you by our season's presenting sponsor, WorkOS.

6:17

SPEAKER_00

What do OpenAI, Anthropic, Cursor, Vercel, Replit, Sierra, Clay, and hundreds of other winning companies all have in common?

6:21

SPEAKER_01

[SPEAKER_00] They are all powered by WorkOS.

6:25

SPEAKER_00

If you're building a product for the enterprise, you've felt the pain of integrating single sign-on, SCIM, RBAC, audit logs, and other features required by large companies. WorkOS turns those deal blockers into drop-in APIs with a modern developer platform built specifically for B2B SaaS. Literally every startup that I'm an investor in that starts to expand upmarket ends up working with WorkOS.

6:46

SPEAKER_01

[SPEAKER_00] And that's because they are the best. [SPEAKER_00] Whether you are a seed-stage startup [SPEAKER_00] trying to land your first enterprise customer [SPEAKER_00] or a unicorn expanding globally, [SPEAKER_00] WorkOS is the fastest path to becoming enterprise-ready [SPEAKER_00] and unblocking growth. [SPEAKER_00] It's essentially Stripe for enterprise features. [SPEAKER_00] Visit WorkOS.com to get started, [SPEAKER_00] or just hit up their Slack, [SPEAKER_00] where they have actual engineers [SPEAKER_00] waiting to answer your questions. [SPEAKER_00] WorkOS allows you to build faster [SPEAKER_00] with delightful APIs, comprehensive docs,

7:12

SPEAKER_01

[SPEAKER_00] and a smooth developer experience. Go to WorkOS.com to make your app enterprise-ready today. [SPEAKER_00] So if we're in this 1997 timeline for AI, [SPEAKER_00] I know so much of your message is, [SPEAKER_00] “We don't know where it's going exactly yet.” [SPEAKER_00] Do you have a sense of the timeline [SPEAKER_00] to, okay, now things are going to be radically changing? [SPEAKER_00] Where are we in that cycle? [SPEAKER_00] You talk about all these different cycles we've been through. [SPEAKER_00] How far are we from, “Wow, it's all different now”? Well, unquestionably, we're already in that moment in software. And then there's a conversation about,

7:38

SPEAKER_01

well, what does agentic and AI software development, two separate things that merge together, mean for the future of the software industry? There's one extreme, which no one really believes, which is, hey, you'll just VibeCode your own Stripe. And no one actually believes that, although prove you don't believe that, but clearly there's a whole bunch of questions about what this means for the software industry and how much stuff you'll be able to do yourself or how much more software there will be. And that's one whole conversation. But the other extreme is, if you're in a law firm, this is all very interesting, but what am I— how exactly do we use this,

8:11

SPEAKER_01

and how do we work out how not to be the next story that we've submitted something with hallucinations in it? And how many associates are we going to hire next year? What does this mean for us? One of the analogies I used in the presentation is: Imagine you're seeing— imagine you're an accountant seeing the first software spreadsheets in the late '70s. And this is mind-blowing. Now you change the interest rate here, and all the other numbers change. And it does a week of work for you in 30 seconds. And we can talk about what that meant for the accounting industry. But clearly, if you're an accountant, this is obviously mind-blowing.

8:51

SPEAKER_01

But if you were a lawyer looking at that or a journalist looking at that, you'd think, “Well, that's very clever, and my accountant should see this, but that's not what I do. I might use it for my timesheet next week if it didn't cost $10,000 or $15,000 to get the Apple II and the monitor and the printer to run it,” which is what it costs if you adjust for inflation. But that's not what I do. And you need a word processor, which actually came very shortly afterwards. And so that's the moment that we're in. Some people, such as software developers, are the accountancy in VisiCalc. “Oh my God, this changes everything.” Before VisiCalc and after VisiCalc,

9:28

SPEAKER_01

before Claude Code and after Claude Code. A lot of other people are picking it up, using it to varying degrees, but slightly puzzled. So there's a bunch of survey data that I put in the presentation that even if you look at 13- to 18-year-olds or something, it's still 15%, 20% of people who are daily active users,

9:45

SPEAKER_00

[SPEAKER_01] and another 20% are weekly active users. [SPEAKER_01] are the accountancy in VisiCalc. [SPEAKER_01] “Oh my God, this changes everything.” [SPEAKER_01] Before VisiCalc and after VisiCalc, [SPEAKER_01] before Claude Code and after Claude Code. [SPEAKER_01] A lot of other people are picking it up, [SPEAKER_01] using it to varying degrees, [SPEAKER_01] but slightly puzzled. [SPEAKER_01] So there's a bunch of survey data [SPEAKER_01] that I put in the presentation [SPEAKER_01] that, even if you look at 13- to 18-year-olds [SPEAKER_01] or something, it's still 15%, 20% of people [SPEAKER_01] who are daily active users,

10:04

SPEAKER_00

[SPEAKER_01] and another 20% are weekly active users.

10:06

SPEAKER_01

And then the other 60% of those people in that demographic, how long do you say they are not using this? So there's a very wide spread of who gets it and a very wide spread, which I think also maps— this is almost a separate point— to the jagged frontier question of where does this work? Where does it not work? [SPEAKER_00] [SPEAKER_01] Can you tell where it's going to work? [SPEAKER_00] [SPEAKER_01] Is it intuitive to know where it would work? [SPEAKER_00] [SPEAKER_01] Can you tell after it worked? [SPEAKER_00] [SPEAKER_01] Can you work out for yourself [SPEAKER_00] [SPEAKER_01] what you would do with this? [SPEAKER_00] [SPEAKER_01] And all of those intersect.

10:27

SPEAKER_01

[SPEAKER_00] [SPEAKER_01] If you're a software developer, [SPEAKER_00] [SPEAKER_01] a lot of other people were— [SPEAKER_00] [SPEAKER_01] people are having a moment or they're not, [SPEAKER_00] [SPEAKER_01] or we're in, again, [SPEAKER_00] [SPEAKER_01] we're in that 1997 moment of— [SPEAKER_00] [SPEAKER_01] maps to the jagged frontier question [SPEAKER_00] [SPEAKER_01] of where does this work? [SPEAKER_00] [SPEAKER_01] Where does it not work? Can you tell where it's going to work? Is it intuitive to know where it would work? Can you tell after it worked? Can you work out for yourself what you would do with this? And all of those intersect if you're a software developer.

10:47

SPEAKER_01

A lot of other people were, “People are having a moment or they're not,” or we're in, again, we're in that 1997 moment of, “Okay, what is this?” [SPEAKER_00] Along those lines, [SPEAKER_00] something you've been writing a bit about [SPEAKER_00] is this unexpected investment [SPEAKER_00] in professional services [SPEAKER_00] slash consulting services [SPEAKER_00] slash forward-deployed engineers. [SPEAKER_00] All the AI labs, [SPEAKER_00] at least the two big ones, [SPEAKER_00] OpenAI and Anthropic, [SPEAKER_00] are investing in [SPEAKER_00] buying massive consultancies [SPEAKER_00] and PE firms. [SPEAKER_00] Talk about just what's happening there

11:12

SPEAKER_01

[SPEAKER_00] and why that's happening. Well, I was groping for a joke last night when I wrote my newsletter and couldn't quite get it to land. But as you know, something: You already know the joke that a machine learning scientist is a statistician who lives in San Francisco. And there's something in there: A forward-deployed engineer is an Accenture outsourced software developer who lives in San Francisco or works in San Francisco. Joking apart, if you have any experience of professional services, companies do not have lots of people sitting around waiting to build a big new project, or do a big new piece of analysis, or build a big new piece of technology,

11:45

SPEAKER_01

or a new product, or work out how they're going to redesign their stores, or work out where the stores should be, or try and work out why the churn is too high. All of those kinds of questions are reasons why you hire Bain, BCG, McKinsey on one side, or Accenture, Infosys, whoever on the other, or you hire a branding agency, or you hire a firm of architects, or whatever. And it's always, “Well, we could hire some architects, but why on earth would we want to have

12:11

SPEAKER_00

[SPEAKER_01] 15 architects on staff [SPEAKER_01] when we just go [SPEAKER_01] and hire an architecture firm, [SPEAKER_01] and we just go [SPEAKER_01] and hire an ad agency?” [SPEAKER_01] And so you're supposed to [SPEAKER_01] completely reimagine [SPEAKER_01] all of the internal workflows [SPEAKER_01] of your company [SPEAKER_01] and work out [SPEAKER_01] which of them [SPEAKER_01] could be automated

12:29

SPEAKER_01

really quickly with AI. That's a project. That's a project that needs five or ten people to sit down and spend a month or two working it out, and then actually doing it is another project. Okay, so you need to plug [SPEAKER_00] [SPEAKER_01] these three vertical systems [SPEAKER_00] [SPEAKER_01] into these two horizontal systems, [SPEAKER_00] [SPEAKER_01] and build a bunch of new workflows, [SPEAKER_00] [SPEAKER_01] and train people to do that. [SPEAKER_00] [SPEAKER_01] Well, guess what? [SPEAKER_00] [SPEAKER_01] Who's going to do that? [SPEAKER_00] [SPEAKER_01] Because you don't have [SPEAKER_00] [SPEAKER_01] a bunch of people [SPEAKER_00] [SPEAKER_01] sitting around

12:54

SPEAKER_01

[SPEAKER_00] [SPEAKER_01] not doing anything. [SPEAKER_00] [SPEAKER_01] So, on the one side, [SPEAKER_00] [SPEAKER_01] into these two horizontal systems, [SPEAKER_00] [SPEAKER_01] and build a bunch of new workflows, [SPEAKER_00] [SPEAKER_01] and train people to do that. [SPEAKER_00] [SPEAKER_01] Well, guess what? [SPEAKER_00] [SPEAKER_01] Who's going to do that? [SPEAKER_00] [SPEAKER_01] Because you don't have [SPEAKER_00] [SPEAKER_01] a bunch of people [SPEAKER_00] [SPEAKER_01] sitting around [SPEAKER_00] [SPEAKER_01] not doing anything. [SPEAKER_00] [SPEAKER_01] So, on the one side, [SPEAKER_00] [SPEAKER_01] this is part of the model of some PE firms,

13:18

SPEAKER_01

which is that they provide support to their portfolio companies to do stuff. And on the other side, that's why you hire— depending on what you're trying to do— you hire Bain, or you hire Accenture, or you hire Publicis [SPEAKER_00] to help you work that out. [SPEAKER_00] What's really funny [SPEAKER_00] about this trend [SPEAKER_00] is you would think [SPEAKER_00] consultants were going to be gone. [SPEAKER_00] No, we don't need [SPEAKER_00] all these people anymore. [SPEAKER_00] AI is going to do their work. [SPEAKER_00] Instead, the most cutting-edge AI labs [SPEAKER_00] are the ones most investing [SPEAKER_00] in these folks.

13:40

SPEAKER_01

[SPEAKER_00] I think it's pretty surprising. Well, one of the strands in my presentation— so I split the presentation into three sections. There's a section on capital, which is: Where is all this capex going, and are the model labs going to have differentiation? And then there's a section on deployment, which is: What does it mean for the software industry? And then the third section is: How does this change stuff? And one of the strands I tried to pull together in the section on change is: What's the hard part of the job? Is the hard part of the job writing the code line by line? Is the hard part of the job giving you the SKU or making the PowerPoint?

14:08

SPEAKER_01

Or is the hard part of the job something else? Is it the task or the job? And pulling that apart, sometimes the task is the job. The classic example is an elevator attendant. If I live in a building that has an attended elevator, we have a manual elevator. There's no button. There's a lever, and the doorman drives you to your floor. It's a vertical speed car. Giving you the SKU or making the PowerPoint? Or is the hard part of the job something else? Is it the task or the job? And pulling that apart, sometimes the task is the job. The classic example is an elevator attendant. If I live in a building that has an attended elevator, we have a manual elevator.

14:35

SPEAKER_01

There's no button. There's a lever, and the doorman drives you to your floor. It's a vertical streetcar. It's like one of those trams in San Francisco. They drive you to your floor. And then those all got automated after the '50s, and now you get in and you press a button, and pressing the button is the job. So there were some things where the job was a task, and the task got automated. What happens much more, and this is why people talked about the Jevons paradox, is this price elasticity, because Jevons paradox is just price elasticity, applied price elasticity. If you make it cheaper to do something, what happens? Do you do the same for less money, or do you do more

15:16

SPEAKER_01

for the same amount of money, or do you do more for more money because you've got a new ROI? And if you look at something like the history of accounting, or indeed professional services, this is a joke I made on Twitter back when it was Twitter: Young people won't believe this, but before Excel, junior investment bankers worked really long hours, and now, thanks to Excel, Goldman's associates all leave work at lunchtime on Fridays. Well, why is that not what happened? You could make the same point about software development. Before IDEs, libraries, and operating systems, developers had to write all the code. Now, if you write an iPhone app, 90% of the code

15:47

SPEAKER_01

is written for you by Apple. Apple wrote the modem driver, the graphics drivers, and the file system. You don't need to write any of that. So we've got a tenth as many engineers now. Well, no. And so then you have to look at an industry and work out, well, which is it, and what is the hard part? One of the analogies An iPhone app: 90% of the code is written for you by Apple. Apple wrote the modem driver, the graphics drivers, and the file system. You don't need to write any of that. So we've got a tenth as many engineers now. Well, no. And so then you have to look at an industry and work out, well, which is it, and what is the hard part? One of the analogies

16:22

SPEAKER_01

that occurred to me here is to look at the history of e-commerce, which is that what Amazon does is it gets you the SKU if you know what the SKU is. If you know what SKU you want— you want that microphone stand, this part number— you can go to Amazon and get it. If you don't know what microphone to get, you probably shouldn't start on Amazon. Multiply that by many, many, many product categories. And so what Amazon does is get you the SKU, but knowing what SKU you want is another job. Claude Code can write you the code, but what code do you want? It can make you the features, sure, but what features do you want? Who's your customer? What's the right product

16:49

SPEAKER_01

for that customer? How are you going to take it to market? And, a long way of answering the question: Why do you hire McKinsey? Are you hiring them to get a 75-slide deck? Well, narrowly, Claude Cowork will make a really, really crappy version of that. And you'll get all these AI grifters on LinkedIn, Twitter, and so on, saying, "Hey, I made a McKinsey deck with Claude," and you look at it and you think, "Yeah, that's a bunch of dog crap. That's not what you get from McKinsey." But even if it was, that's not what you pay them for. What you actually pay Bain to do is to go and walk all over your company and work out, yes, but why is it that you didn't do that?

17:26

SPEAKER_01

And how do the politics of this work? And what do you actually need to do? And let's go and talk to your customers and work out what they actually think, as opposed to what's on the first page of Google. It's all the other stuff, and the PowerPoint is just the task, but that's not what you hired them for. The same with Amazon versus the retailer,

17:45

SPEAKER_00

[SPEAKER_01] the same with software [SPEAKER_01] development. [SPEAKER_01] So you've got that. [SPEAKER_01] Need to do? [SPEAKER_01] And let's go and talk [SPEAKER_01] to your customers [SPEAKER_01] and work out [SPEAKER_01] what they actually think, [SPEAKER_01] as opposed to [SPEAKER_01] what's on the first page [SPEAKER_01] of Google. [SPEAKER_01] It's all the other stuff, [SPEAKER_01] and the PowerPoint [SPEAKER_01] is just the task, [SPEAKER_01] but that's not [SPEAKER_01] what you hired them for. [SPEAKER_01] The same with [SPEAKER_01] Amazon versus the retailer, [SPEAKER_01] the same with software [SPEAKER_01] development. [SPEAKER_01] So you've got that

18:05

SPEAKER_00

[SPEAKER_01] split. [SPEAKER_01] The other analogy [SPEAKER_01] that occurred to me here [SPEAKER_01] is looking at

18:09

SPEAKER_01

the class of industry that got steamrolled by the internet because they had those two things, and you could split them apart. So you had the physical manufacturing or physical distribution, and then you had the other, the thing. What was the actual thing? Classic examples would be newspapers and recorded music. So record companies do not think of themselves as being in the business of manufacturing small pieces of plastic, but that was, that was what they actually did, [SPEAKER_00] [SPEAKER_01] and when that went away, [SPEAKER_00] [SPEAKER_01] they were screwed. [SPEAKER_00] [SPEAKER_01] Same thing for newspapers. [SPEAKER_00] [SPEAKER_01] Newspapers did not think

18:38

SPEAKER_01

[SPEAKER_00] [SPEAKER_01] of themselves as [SPEAKER_00] [SPEAKER_01] manufacturing and trucking companies. [SPEAKER_00] [SPEAKER_01] When you decouple that, [SPEAKER_00] [SPEAKER_01] then that becomes a problem, [SPEAKER_00] [SPEAKER_01] but often you [SPEAKER_00] [SPEAKER_01] can't decouple that, [SPEAKER_00] [SPEAKER_01] or that wasn't really [SPEAKER_00] [SPEAKER_01] the problem, [SPEAKER_00] [SPEAKER_01] or you make that thing cheap, [SPEAKER_00] [SPEAKER_01] and then all this other stuff [SPEAKER_00] [SPEAKER_01] happens as well. [SPEAKER_00] [SPEAKER_01] And so all of this [SPEAKER_00] [SPEAKER_01] is just vastly more complicated

18:51

SPEAKER_01

[SPEAKER_00] [SPEAKER_01] than saying, [SPEAKER_00] [SPEAKER_01] "Well, hey, [SPEAKER_00] [SPEAKER_01] we're just going to [SPEAKER_00] [SPEAKER_01] automate the accountants." [SPEAKER_00] [SPEAKER_01] or that wasn't really [SPEAKER_00] [SPEAKER_01] the problem, [SPEAKER_00] [SPEAKER_01] or you make that thing cheap, [SPEAKER_00] [SPEAKER_01] and then all this other stuff [SPEAKER_00] [SPEAKER_01] happens as well. [SPEAKER_00] [SPEAKER_01] And so all of this [SPEAKER_00] [SPEAKER_01] is just vastly more complicated [SPEAKER_00] [SPEAKER_01] than saying, [SPEAKER_00] [SPEAKER_01] “Well, hey, [SPEAKER_00] [SPEAKER_01] we're just going to

19:05

SPEAKER_01

[SPEAKER_00] [SPEAKER_01] automate the accountants, [SPEAKER_00] [SPEAKER_01] or we're going to [SPEAKER_00] [SPEAKER_01] automate the consultants.” [SPEAKER_00] [SPEAKER_01] There are two charts [SPEAKER_00] [SPEAKER_01] in the presentation of the number of people employed as accountants, which went up right the way through the 20th century and has gone up again since the beginning of the 21st century. So you have adding machines and punch cards and mainframes and databases and ERP and cloud spreadsheets and PCs, and the number of accountants keeps going up. And so why is that? Well, it's not— it must be more— it's more complicated than automation.

19:35

SPEAKER_01

[SPEAKER_00] Even just looking at [SPEAKER_00] the most advanced [SPEAKER_00] AI companies [SPEAKER_00] throughout Big OpenAI. [SPEAKER_00] I just had Dan Shipper [SPEAKER_00] from Every on the podcast. [SPEAKER_00] Everyone's just [SPEAKER_00] increasing headcount. [SPEAKER_00] The companies [SPEAKER_00] you would think [SPEAKER_00] would be least [SPEAKER_00] likely to add humans [SPEAKER_00] are adding [SPEAKER_00] many, many humans. [SPEAKER_00] And to your point, [SPEAKER_00] it's really complicated. [SPEAKER_00] What's your—just [SPEAKER_00] on the job, [SPEAKER_00] the coming jobpocalypse? [SPEAKER_00] Dario's talking [SPEAKER_00] about all the

19:52

SPEAKER_01

[SPEAKER_00] entry-level people [SPEAKER_00] or no more jobs, [SPEAKER_00] just... Yeah. There's a narrow point here, which is that I would place... I don't like arguments from authority. And I don't think the fact that you run an AI lab suddenly gives you... Or rather, if you're going to use arguments from authority, then they should be relevant to the field. So I'm interested in Dario's opinions on where models are going to go in the next 6 to 12 months. Not particularly interested in opinions on theories of labor, market value, and comparative advantage. Yeah, maybe he had a course on that at university. So did I. So I think one needs to be a little bit cautious

20:26

SPEAKER_01

about what Dario says. And that's setting aside the cynical view that he's just doing that to pump stock, which I don't believe at all. So it comes back to my point about platform shifts. Every time we have a new technology, it automates away a bunch of jobs. And then that automation, whether it's price elasticity and the enablement of the fact that they became automated, unlocks a bunch of new jobs. And so, you go back to 1800: 90% of us were peasants. And our major concern was, “Are the crops going to fail?” Because then we'll all go hungry, or worse. And so ever since then, we've been automating jobs and creating new jobs. And you can always see the job that's going

20:55

SPEAKER_01

to go away. And you don't know the new job because it doesn't exist yet. And it's something that sounds dumb anyway. Railway engineer. What's a railway? Why would that be a thing? Who would want to go that fast? And so we've had that process over and over again. This is what any first-year economics student would tell you. We've had this process over and over again since 1800. And you don't know the new job because it doesn't exist yet. And it's something that sounds dumb anyway. Railway engineer. What's a railway? Why would that be a thing? Who would want to go that fast? And so we've had that process over and over again. This is what any first-year economics student

21:23

SPEAKER_01

would tell you. We've had this process over and over again since 1800. And each time you go through it, you get a bunch of frictional pain and dislocation, and a bunch of people lose their jobs, and a bunch of towns get hollowed out, and it all sucks. But when you come through on the other side, we're all richer, and we're not worried about the crops failing anymore. And this is the process of the last 200 years. So then the question is: Is there some a priori reason why this would be different from those? Because the internet removed a bunch of jobs. PCs removed a bunch of jobs. There aren't many people working as typesetters anymore, or telephone operators, or typists.

21:45

SPEAKER_01

The internet removed a bunch of jobs, [SPEAKER_01] and generally the jobs that go away are crap jobs, seen retrospectively, and the new jobs are better because GDP keeps going up. So is AI different? And so then there are a couple of answers to this. One theory is: Well, this is going to be way quicker. And certainly the adoption of AI is quicker than previous technologies. But this is because you're standing on the shoulders of giants. So you don't need to wait for everyone to buy a piece of expensive hardware— to buy a phone or a PC— or wait for the telco to deploy broadband. It's already there. So, of course, ChatGPT can get 900 million WeChat users

22:07

SPEAKER_01

because there are already 900 million people on the internet. When Marc Andreessen launched Netscape in—what was it?— '93, '94, there were 50 to 100 million PCs on Earth. So, no, you didn't have 900 million users then. But the point is, then, he didn't need microchips. And before that, you didn't need to wait for electricity, and you didn't need to wait for mass production. So you're always standing on the shoulders of giants. There's always a compounding effect. So, yeah, this is faster, but the internet was faster too. I think the other answer to this, and this comes back to the professional services point, is: You talk to these doomers on Twitter, and they would

22:40

SPEAKER_01

act as if every big company is going to buy ChatGPT tomorrow, and then in two weeks' time they'll fire all their staff, and these people are morons. This is one of many reasons why the doomers were morons: a complete failure to understand the way the world works. And that was the starting point for why they then didn't understand anything else. A typical big-company enterprise software sales cycle— you'll know this better than me— is 18 months if you're lucky. This is always a problem. The enterprise sales cycle is shorter than the venture-backed startup funding cycle. Longer, rather. Longer.

23:08

SPEAKER_01

It takes you longer to get an enterprise deal than it takes you to go between rounds. And this was always a problem, particularly for sectors like aerospace or healthcare or something. So, I know people aren't going to tear out SAP and replace it with X, Y, Z. Maybe in three, five, ten years, yes, that whole estate will look radically different, and all those jobs will have changed. But it will take two, three, four, five, ten years, and it will take time sector by sector, and it will take time for people to work out, "Oh, you could do that thing with this."

23:09

SPEAKER_01

And one of the companies I always remember that we looked at when I was at Andreessen Horowitz is a company called Frame.io, which is video editing, video collaboration. And there's nothing there that you couldn't have done at least five years earlier, and maybe ten years earlier. And actually, that's a bad example because that relies on a bunch of stuff—cutting-edge web technologies.

23:10

SPEAKER_01

So, I know people aren't going to tear out SAP and replace it with X, Y, Z. Maybe in three, five, ten years, yes, that whole estate will look radically different, and all those jobs will have changed. But it will take two, three, four, five, ten years, and it will take time sector by sector, and it will take time for people to work out, "Oh, you could do that thing with this."

23:10

SPEAKER_01

And one of the companies I always remember that we looked at when I was at Andreessen Horowitz is a company called Frame.io, which is video editing, video collaboration. And there's nothing there that you couldn't have done at least five years earlier, and maybe ten years earlier. And actually, that's a bad example because that relies on a bunch of stuff—cutting-edge web technologies.

23:10

SPEAKER_01

If you go out and pick ten random SaaS companies that were started the day before ChatGPT launched, how many of them could have been founded at any point in the previous 15 years? The delay was somebody realizing, "Oh, we could—that problem exists inside that industry, and, oh, this is the way that we would solve it." It didn't all happen the day after Google Docs. It took ten, fifteen, twenty years for people to invent all that stuff and work out that you could do that with this. And so all of that is the way of saying, well, yes, it is going to be quick, but actually, no, it will take a while for people to work out how to completely change how their business works.

23:11

SPEAKER_01

[SPEAKER_00] Your view is so comforting because it's, okay, this is a huge deal, but we've been through many transformations before, and it's going to be okay. Well, I have a slide towards the end of the presentation, which—I know the title is something—"This is going to be completely different from everything else, just like everything else."

23:13

SPEAKER_01

And then the next slide is an IBM ad from the '50s, which has this sea of white men in white shirts and ties, all holding up slide rules. And the slogan in the title of the ad is—it's an IBM ad—it says, "An IBM electronic calculator"—this is before it was called a computer; it's an electronic calculator, it's the size of a fridge—"It's like having 150 extra engineers."

23:13

SPEAKER_01

How many people listening to this come from companies whose slogan is, "We'll give you 150 extra engineers"? Isn't that the whole pitch of Claude Code? 150 extra engineers for free—or not free; that's a lot of money. And yeah, that's what it gave you. And so, yes, we keep going through this over and over and over again. Comfort list—their company's slogan is, “We’ll give you 150 extra engineers.” Isn't that the whole picture of Claude Code? One hundred and fifty extra engineers for free—or not free; that's a lot of money. And yeah, that's what it gave you. And so, yes, we keep going through this over and over and over again, just to make that tangible.

23:14

SPEAKER_01

Obviously, we couldn't be doing this without the internet. There's a slide in my presentation that we could maybe talk about. It's a slide or chart showing how many products have been stocked in supermarkets in America since the ’50s. And the point of the slide is to say that barcodes allowed supermarkets to stock way more stuff because they could keep track of it.

23:15

SPEAKER_01

But to make that chart, I had to know there was a thing called the Food Marketing Institute. And I had to have found out that they published a number for how many SKUs there were in supermarkets every year. And then I had to realize they'd been around since the ’50s. And if I dug long enough, I might be able to make a whole time series, and I could make a whole chart.

23:16

SPEAKER_01

Now imagine doing that in 1994. First of all, you would have no idea that exists. You'd really need to go and find a library where they published that number and know that the number's in that report. You'd have no idea. Then you'd need to find a library that had them. So you're going to spend three days on the phone and spend $50 on long-distance phone calls to find a library that has these. Or maybe you call the Food Marketing Institute and they say, “Yeah, sure, we'll sell them to you for $500 each.” So then you're going to go on a trip. Maybe you live in New York or somewhere that has this. And two weeks later, you've got the chart, and you look at it.

23:17

SPEAKER_01

And then the other side of this is that the life of an analyst is: You spend all day making a chart, and you look at it and go, “Oh, that's not really interesting.” So you spend two weeks making the chart, and then you look at it and go, “Yeah, I'm not going to use that.” And for me, this was two hours on Google.

23:18

SPEAKER_00

[SPEAKER_01] And so we forget how big a deal the internet was. That's a long way of saying it, but we forget we've had these absolutely enormous changes, and then we don't see it because that's the world— [SPEAKER_01] [SPEAKER_00] The world's always been. What's potentially different this time, even though your quote is, “It's different. Everything's going to change just like—just like last time”? The big difference, obviously, is AGI might emerge, and superintelligence, where it could do the work humans can do, do a lot of this stuff for us, and actually replace jobs. Just thoughts on that element of this transformation we're going through.

23:22

SPEAKER_00

[SPEAKER_01] I don't know. This is one of the ways I've struggled to write about AI. Certainly, in 2023 and early 2024, all the questions were questions you could have asked in December 2022. The questions didn't really change, and the strategies didn't really change. And I think the AGI question is the same.

23:22

SPEAKER_00

[SPEAKER_01] The observation one can make is that we have no theory of what human intelligence is. We have no theory of why these models work so well. We have no theory of how much better they will get. So we're all just vibes forecasting as to what will happen. And then you can have the 2 a.m. doped-out philosophy students talking about, “Hey, man, is this consciousness? Maybe we aren't conscious either. We just think we are.” Yeah, great. Thank you. I think the one thing—

23:23

SPEAKER_00

[SPEAKER_01] We have no theory of how much better they will get. So we're all just vibes forecasting as to what will happen. And then you can have the 2 a.m. doped-out philosophy students talking about, “Hey, man, is this consciousness? Maybe we aren't conscious either. We just think we are.” Yeah, great. Thank you. [SPEAKER_01] I think the one thing one can observe today is that we have no idea. We don't know. We can guess, but we don't really know where this is going to end up. What I think you can say today is that there's a lot of redefinition of terms.

23:24

SPEAKER_00

[SPEAKER_01] A quote I used in my presentation late last year was from an AI scientist called Larry Tesler, who said, “AI is whatever machines can't do yet,” because once machines can do it, people say, “Well, that's just software.” [SPEAKER_01] And so, certainly, I did a poll on social media every now and then asking, “Is machine learning still AI?” Because I've certainly heard people say, “Oh, that's not AI. That's just image recognition. That's not AI. That's just sentiment analysis.”

23:26

SPEAKER_00

[SPEAKER_01] So AI is a bit like the word “technology.” If it's new, then it's technology. But in the ’60s, jet airliners were technology. Now a jet airliner isn't tech. And so there's a sense of AI as a moving target: whatever just started working.

23:28

SPEAKER_01

And I think the point here is that now, clearly, you can see people redefining AGI to mean the stuff that works now. So is AGI—what's the definition now? It can do a certain percentage of economically valuable work. Well, that's a very different thing from “It has a soul and it's fucking alive,” because a database could do that. An IBM mainframe in 1975 could do a meaningful percentage of economically valuable work that was previously done by people. And it turned out there was a whole bunch of other stuff that it couldn't do, that we didn't do then. We didn't know it existed.

23:28

SPEAKER_01

And I think the point here is that now, clearly, you can see people redefining AGI to mean the stuff that works now. So is AGI—what's the definition now? It can do a certain percentage of economically valuable work. Well, that's a very different thing from “It has a soul and it's fucking alive,” because a database could do that. An IBM mainframe in 1975 could do a meaningful percentage of economically valuable work that was previously done by people. And it turned out there was a whole bunch of other stuff that it couldn't do, that we didn't do then. We didn't know it existed.

23:28

SPEAKER_01

So there's a lot of creative redefinition here. Superintelligence—I'm not sure, is superintelligence more than AGI or less than AGI? Because last year, I thought superintelligence was really good, but not as good—not actual AGI. And now it's, “Oh, no, no, we've already got AGI. But superintelligence, that's really hard.” [SPEAKER_00] It's all these terms.

23:30

SPEAKER_01

What are you—I didn't even—what even? It's funny, I was having an argument on Hacker News this morning. We remember the idea—you remember the argument—which is never a good use of time, but you remember the argument of people arguing about whether crypto is blockchain or whether blockchain is crypto. There isn't a right answer to that. Let's just be clear: it's important to understand what you mean when you say that, but there isn't a correct answer to this. Are we going to get to something that has human-level intelligence? I don't really know. I don't think we have any way of answering that question. Maybe, maybe not. You can make arguments either way.

23:32

SPEAKER_01

Meanwhile, we've got this thing that's clearly a completely transformative technology. And maybe the serious point here is that you don't have to believe—even if the models stopped getting better tomorrow, if this is it and we hit a brick wall tomorrow, this is an incredibly useful technology that's going to change the world and get rolled out over the next 10 years. So you don't have to believe in any of that stuff to believe that this is a giant deal.

23:33

SPEAKER_01

[SPEAKER_00] Something that's definitely changed: I had my former boss, Marc Andreessen, on the podcast. And we didn't actually talk about this during the conversation. He brought it up before we started recording, and I never got to it. He had this insight that the opportunity set for companies now is so much larger. Over the next 10 years. So you don't have to believe in any of that stuff to believe that this is a giant deal. [SPEAKER_00] Something that's definitely changed: [SPEAKER_00] I had the former boss, [SPEAKER_00] Marc Andreessen, [SPEAKER_00] on the podcast. [SPEAKER_00] And we didn't actually talk about this [SPEAKER_00] during the conversation,

23:44

SPEAKER_01

[SPEAKER_00] and he brought it up [SPEAKER_00] before we started recording, [SPEAKER_00] and I never got to it. [SPEAKER_00] He had this insight [SPEAKER_00] that the opportunity set [SPEAKER_00] for companies now [SPEAKER_00] is so much larger. [SPEAKER_00] We used to have [SPEAKER_00] no trillion-dollar companies. [SPEAKER_00] Now we're going to have [SPEAKER_00] dozens of trillion-dollar companies. [SPEAKER_00] Just the size [SPEAKER_00] companies can grow to [SPEAKER_00] is going up so much. [SPEAKER_00] And valuations also go up [SPEAKER_00] along with that. [SPEAKER_00] And his point is just [SPEAKER_00] people haven't really grokked

24:04

SPEAKER_01

[SPEAKER_00] just how large companies [SPEAKER_00] can get now. [SPEAKER_00] Everyone's hitting [SPEAKER_00] a hundred million ARR [SPEAKER_00] in five months, [SPEAKER_00] six months. [SPEAKER_00] Just thoughts on that. Yeah. This was his whole “software is eating the world” thesis from 15 years ago, whenever it was. Yeah. The TAM gets progressively bigger because you can address [SPEAKER_00] [SPEAKER_01] larger and larger parts [SPEAKER_00] [SPEAKER_01] of the economy. [SPEAKER_00] [SPEAKER_01] And so, [SPEAKER_00] [SPEAKER_01] if you think about [SPEAKER_00] [SPEAKER_01] the classic [SPEAKER_00] [SPEAKER_01] platform-shift framing,

24:24

SPEAKER_01

[SPEAKER_00] [SPEAKER_01] mainframes are— [SPEAKER_00] [SPEAKER_01] I think peak mainframe [SPEAKER_00] [SPEAKER_01] install base was something like 70,000, 80,000 units. A slightly fuzzy term: What exactly is a mainframe, and what's the difference? At what point does it become two mainframes or one? But something like that, that order of magnitude. And then when the internet kicks off, there are, as I said, 50 to 100 million PCs on Earth. Maybe today there are something over a billion, one to one and a half billion, but obviously a lot of those are corporate. It's 700 million, 800 million consumer PCs in the world. There's about five and a half,

24:53

SPEAKER_01

six billion mobile smartphones in the world, which is why you can have 900 million weekly active users on ChatGPT. And so there was this narrative five years ago: We've run out of people. So the next thing can't be an order of magnitude bigger, which was true up to a point, but that was the wrong model because clearly what's happening now is you're moving in another direction: You're just branching out and automating big, big new swathes of the economy. Now, back to your job point, you could argue, we're just going to replace all the people with AI, and all the money will go to Sam Altman. So the next thing can't be an order of magnitude bigger,

25:20

SPEAKER_01

which was true up to a point, but that was the wrong model because clearly what's happening now is you're moving in another direction: You're just branching out and automating big, big new swathes of the economy. Now, back to your job point, you could argue, we're just going to replace all the people with AI, and all the money will go to Sam Altman, and Marc can buy himself another Gulfstream. I think the— add to the fleet— I think the other answer is: It's back to the lump of labor fallacy and the last 200 years. Each of these technologies removes a bunch of jobs, creates a bunch of new jobs, creates a bunch of new value, unlocks prosperity for all of us.

25:45

SPEAKER_01

And that's painful as you go through it, but it always creates more value. And so here you could, you could certainly make an analogy to the useful analog of the electricity industry, just saying how electricity became part of absolutely everything. And software has been slowly working its way out. The analog here

26:02

SPEAKER_00

[SPEAKER_01] would be electricity [SPEAKER_01] in factories, [SPEAKER_01] and then electricity [SPEAKER_01] slowly spreads out. [SPEAKER_01] And so that would be [SPEAKER_01] the point again: [SPEAKER_01] It slowly spreads out [SPEAKER_01] to do more and more things. [SPEAKER_01] And so, [SPEAKER_01] more and more value [SPEAKER_01] and a bigger and bigger [SPEAKER_01] contribution to the economy. [SPEAKER_01] It also, [SPEAKER_01] of course, [SPEAKER_01] disappears inside things. [SPEAKER_01] And the other side, [SPEAKER_01] the point of my capital section [SPEAKER_01] in the presentation, [SPEAKER_01] is there's this quote [SPEAKER_01] from Sam Altman

26:21

SPEAKER_00

[SPEAKER_01] where he said, [SPEAKER_01] “We're going to be selling [SPEAKER_01] electricity. [SPEAKER_01] We're going to be selling [SPEAKER_01] AI intelligence

26:27

SPEAKER_01

on a meter.” Contribution to the economy. It also, of course, disappears inside things. And the other side, the point of my capital section in the presentation, is there's this quote from Sam Altman where he said, “We're going to be selling electricity. We're going to be selling AI intelligence on a meter, like water or electricity.” And you look at this and think, “My dear sweet child, you need me to explain the marginal structure of the utility industry to you.” Because guess what? When you watch television, the TV company isn't paying a percentage of your monthly bill to the electricity company. When you wash your clothes, Bosch isn't paying a percentage of the price

27:01

SPEAKER_01

of the washing machine. And clearly, the much more specific tactical question at the moment is: Do we even end up with three giant models, or does it become hundreds of models, open models, local models, and so on? And even if we do end up with, say, pick a number, three to six to ten giant foundation models [SPEAKER_00] [SPEAKER_01] that cost hundreds [SPEAKER_00] [SPEAKER_01] of billions of dollars [SPEAKER_00] [SPEAKER_01] a year, [SPEAKER_00] [SPEAKER_01] fine, [SPEAKER_00] [SPEAKER_01] do they get all [SPEAKER_00] [SPEAKER_01] the value from that? [SPEAKER_00] [SPEAKER_01] Now, [SPEAKER_00] [SPEAKER_01] I started my career

27:29

SPEAKER_01

[SPEAKER_00] [SPEAKER_01] as a telecoms analyst, [SPEAKER_00] [SPEAKER_01] and so [SPEAKER_00] [SPEAKER_01] I still pay attention [SPEAKER_00] [SPEAKER_01] to it a bit. [SPEAKER_00] [SPEAKER_01] The global mobile industry [SPEAKER_00] [SPEAKER_01] has revenue [SPEAKER_00] [SPEAKER_01] of about a trillion dollars [SPEAKER_00] [SPEAKER_01] a year, [SPEAKER_00] [SPEAKER_01] maybe a bit more now. [SPEAKER_00] [SPEAKER_01] And it spends [SPEAKER_00] [SPEAKER_01] about $200 billion [SPEAKER_00] [SPEAKER_01] a year on CapEx [SPEAKER_00] [SPEAKER_01] every year. [SPEAKER_00] [SPEAKER_01] Total telecoms [SPEAKER_00] [SPEAKER_01] is about 300;

27:43

SPEAKER_01

[SPEAKER_00] [SPEAKER_01] mobile is about 200. [SPEAKER_00] [SPEAKER_01] About 15% to 20% of revenue every year. And if you look at a chart of mobile data consumption, it's an exponential curve, a perfect curve going straight up. And the number now, I think, is about 1,500 to 2,000 times what it was in 2010 globally. And the stocks have gone nowhere in 25 years because it's an ex-growth, low-margin commodity utility where they're selling this objectively amazing piece of global technology infrastructure that has enormous complexity A perfect curve going straight up. And the number now, I think, is about 1,500 to 2,000 times what it was in 2010 globally. And the stocks

28:23

SPEAKER_01

have gone nowhere in 25 years because it's an ex-growth, low-margin commodity utility where they're selling this objectively amazing piece of global technology infrastructure that has enormous complexity and enormous sophistication, but all the cool stuff is made by you. It's made by the people listening to this podcast. It's made by somebody else. This was that pivotal moment where the telcos thought that they would do all the stuff that you did on your iPhone. And not only do they not do it, but Apple doesn't do it either.

28:49

SPEAKER_00

[SPEAKER_01] It's all further up the stack.

28:49

SPEAKER_01

And so this is the elemental question right now around foundation models: Does the model do the whole thing? Can you just go to the chatbot and get the chatbot to do the whole thing? Can the model companies keep building these Claude for X, Claude for Y things, which, to me, look very much like what you see if you hit File > New in Excel? It's the templates, but all of those are actually billion-dollar companies as well. And if not, does it all have to be apps? Quote-unquote, whatever “app” means. And if it all has to be apps, who builds those? They can't all get built by the model labs, just as they didn't all get built by Microsoft. And so, if they're all

29:23

SPEAKER_01

built by other companies, do the foundation models have leverage up the stack the way Windows did? Or is this more like AWS, where, if you're an engineering company or a law firm buying a piece of software, you don't care which cloud it runs on? And you don't have to standardize on AWS because that's where all the software is, and the developers all standardize on AWS

29:46

SPEAKER_00

[SPEAKER_01] because all the customers [SPEAKER_01] use AWS. [SPEAKER_01] That's not how it works. [SPEAKER_01] That's how Windows or iOS [SPEAKER_01] works, [SPEAKER_01] but that's not how cloud works. [SPEAKER_01] And so it does [SPEAKER_01] seem to me: [SPEAKER_01] Buying a piece of software, [SPEAKER_01] you don't care [SPEAKER_01] which cloud it runs on. [SPEAKER_01] And you don't have [SPEAKER_01] to standardize [SPEAKER_01] on AWS [SPEAKER_01] because that's [SPEAKER_01] where all the software is, [SPEAKER_01] and the developers [SPEAKER_01] all standardize on AWS [SPEAKER_01] because all the customers [SPEAKER_01] use AWS. [SPEAKER_01] That's not how it works.

30:10

SPEAKER_00

[SPEAKER_01] That's how Windows or iOS [SPEAKER_01] works, [SPEAKER_01] but that's not how cloud works. [SPEAKER_01] And so it does [SPEAKER_01] seem to me [SPEAKER_01] that if, [SPEAKER_01] if the chatbot [SPEAKER_01] isn't the UX [SPEAKER_01] and it needs to be apps, [SPEAKER_01] and the model companies [SPEAKER_01] aren't going to build that,

30:23

SPEAKER_01

and the models themselves are commodities, at least as you can see them as users, then why would the model companies have pricing power? And wouldn't all the value be further up the stack? Aren't you— have you got three to six companies selling a commodity [SPEAKER_00] [SPEAKER_01] at marginal cost? Now, obviously the SemiAnalysis guys are: no, no, no, there's going to be infinite pricing power forever. I'm sorry, I'm exaggerating. But I think it's really important to draw a distinction between where we are now, where you have radical price disequilibrium, and you've got these— what's the guy? The OpenClaw guy spent one and a half million dollars on tokens last month.

30:55

SPEAKER_01

But that's somebody getting a 50-grand mobile data bill in 2010. That's temporary. What is the steady-state equilibrium point where all of these lines, the lines on the chart, get lined up and we don't have this weird, crazy stuff going on? And then, will you have pricing power, or have you got three or four or five companies all selling the same thing? And so then you should have a pricing— price— you should have lower pricing and lower margins, and the value up the stack. [SPEAKER_00] I am so excited [SPEAKER_00] to tell you about [SPEAKER_00] this season's [SPEAKER_00] supporting sponsor, And then, will you have pricing power, or have you got three or four or five

31:30

SPEAKER_01

companies all selling the same thing? And so then you should have a pricing— price— you should have lower pricing and lower margins, and the value up the stack. [SPEAKER_00] I am so excited [SPEAKER_00] to tell you about [SPEAKER_00] this season's [SPEAKER_00] supporting sponsor, [SPEAKER_00] Vanta. [SPEAKER_00] Vanta helps over [SPEAKER_00] 15,000 companies, [SPEAKER_00] such as Cursor, [SPEAKER_00] Ramp, [SPEAKER_00] Duolingo, [SPEAKER_00] Snowflake, [SPEAKER_00] and Atlassian, [SPEAKER_00] earn and prove [SPEAKER_00] trust with their customers. [SPEAKER_00] Teams are building [SPEAKER_00] and shipping products [SPEAKER_00] faster than ever, [SPEAKER_00] thanks to AI.

31:54

SPEAKER_01

[SPEAKER_00] But as a result, [SPEAKER_00] the amount of risk [SPEAKER_00] being introduced [SPEAKER_00] into your product [SPEAKER_00] and your business [SPEAKER_00] is higher than [SPEAKER_00] it's ever been. [SPEAKER_00] Every security leader [SPEAKER_00] that I talk to [SPEAKER_00] is feeling the increasing [SPEAKER_00] weight of protecting [SPEAKER_00] their organization, [SPEAKER_00] their business, [SPEAKER_00] and, not to mention, [SPEAKER_00] their customer data. [SPEAKER_00] Because things are [SPEAKER_00] moving so fast, [SPEAKER_00] they are constantly [SPEAKER_00] reacting, [SPEAKER_00] having to guess [SPEAKER_00] at priorities,

32:17

SPEAKER_01

[SPEAKER_00] and having to make do [SPEAKER_00] with outdated solutions. [SPEAKER_00] Vanta automates [SPEAKER_00] compliance and risk management [SPEAKER_00] with over 35 [SPEAKER_00] security and privacy [SPEAKER_00] frameworks, [SPEAKER_00] including SOC 2, [SPEAKER_00] ISO 27001, [SPEAKER_00] and HIPAA. [SPEAKER_00] This helps companies [SPEAKER_00] get compliant fast [SPEAKER_00] and stay compliant. [SPEAKER_00] More than ever before, [SPEAKER_00] trust has the power [SPEAKER_00] to make or break [SPEAKER_00] your business. [SPEAKER_00] Learn more [SPEAKER_00] at Vanta.com [SPEAKER_00] slash Lenny. [SPEAKER_00] And as a listener [SPEAKER_00] of this podcast,

32:38

SPEAKER_01

[SPEAKER_00] you get $1,000 [SPEAKER_00] off Vanta. [SPEAKER_00] That's Vanta.com [SPEAKER_00] slash Lenny. [SPEAKER_00] A really interesting [SPEAKER_00] takeaway here [SPEAKER_00] is that your sense [SPEAKER_00] is, over time, [SPEAKER_00] the foundational model [SPEAKER_00] companies— [SPEAKER_00] Anthropic, [SPEAKER_00] OpenAI, [SPEAKER_00] others— [SPEAKER_00] their margins [SPEAKER_00] will get squeezed. [SPEAKER_00] They will not be [SPEAKER_00] as successful [SPEAKER_00] as they are today. [SPEAKER_00] And the bigger [SPEAKER_00] opportunities are [SPEAKER_00] in the application layer: [SPEAKER_00] the people building [SPEAKER_00] on the models,

32:56

SPEAKER_01

[SPEAKER_00] the wrappers. Yeah, this is a very deterministic thesis, which is: the model companies— crucially, what I said is the models don't seem to have network effects. So there doesn't seem to be a winner-takes-all effect where one of these will run away ahead of the others. So you should have competition indefinitely. If you have competition indefinitely, you don't have primary— really radical differentiation of what the product is— then why would you have pricing power? And meanwhile, if you need to have the models don't seem to have network effects. So there doesn't seem to be a winner-takes-all effect where one of these will run away ahead of the others.

33:33

SPEAKER_01

So you should have competition indefinitely. If you have competition indefinitely, you don't have primary, really radical differentiation of what the product is, then why would you have pricing power? And meanwhile, if you need to have thousands of applications that are all different, built by different people, those can't all be built by the model people. So it should end up looking more like cloud than it looks like Windows. Now, that may be completely wrong. And one of the points I make in the presentation is: Imagine having this conversation about the internet in 1997. What would you have got right? Or indeed, having it about mobile in 2000. You would have missed

34:02

SPEAKER_01

almost all of it. built by different people, those can't all be built by the model people. So it should end up looking more like cloud than it looks like Windows. Now, that may be completely wrong. And one of the points I make in the presentation is: Imagine having this conversation about the internet in 1997. What would you have got right? Or indeed, having it about mobile in 2000. You would have missed almost all of it. You certainly would have said that a has-been PC company from Cupertino would win the whole thing. I know I wouldn't have said that. And a search company with a weird logo. Search? What's that got to do with mobile? No, forget it, you're an idiot.

34:40

SPEAKER_01

So we should presume we don't know. But there are these basic building blocks: Why would they have pricing power? I don't know. I had a— when I was a baby analyst in '99, we went to see a dot-com company in the UK that was trying to do online— selling computer crafts, components online. And they had this whole model and this whole story and the brand and the whole thing. And we went up to see them, and we were on the train back from Birmingham, and this senior banker called David Tate— we were all sitting, talking about it, and Tatey says, "It's a low-margin reseller, one-time sales. You can say dot-com all you like; it's a low-margin reseller." And I think

35:13

SPEAKER_01

the crux of this is they're undifferentiated commodity infrastructure providers. There's a lot of science to it, but there's a lot of science in mobile. What do you pay for a flat-panel screen? There are Nobel Prizes in flat-panel screens. They're still a low-margin commodity. I look forward to being proven wrong, proven wrong, but hey— [SPEAKER_00] that's what it looks like now. [SPEAKER_00] This is great. [SPEAKER_00] So I know you're not an investor. [SPEAKER_00] I know you didn't actually [SPEAKER_00] do investing at A16Z, [SPEAKER_00] even though you worked for A16Z. [SPEAKER_00] Partner. [SPEAKER_00] Partner. [SPEAKER_00] Just sit around

35:36

SPEAKER_01

[SPEAKER_00] and pontificate. [SPEAKER_00] Partner. [SPEAKER_00] Would you— [SPEAKER_00] are there companies [SPEAKER_00] you would invest in? [SPEAKER_00] If there are a couple [SPEAKER_00] of companies you'd invest in now, [SPEAKER_00] are there some on that list, [SPEAKER_00] or categories even? I mentioned briefly that I was an analyst. I was a— I was a sell-side equity analyst. I was not a very good sell-side equity analyst, partly because I was not interested in talking to clients, partly because I was not interested in share prices, which would seem to be a disqualification to being an equity analyst. And I don't— there's, there's, there's a huge difference

36:06

SPEAKER_01

between being right and being early. And there's a huge difference between the right company and the right price. Now, deterministically, you can look across the market and say, it's the bell curve IQ meme, and the guy with 50 [SPEAKER_01] and the guy with 200 [SPEAKER_01] are both saying, "Jeff Bezos, smart guy, I buy stock." And you can certainly overthink all of this. And you can look at Google, Apple, Facebook, Amazon and say, "Hard to see a problem for them, really, with all of this." You can certainly see questions for all of them. And one of them may drop the ball, but it's worth remembering what happened in mobile. The internet was just

36:37

SPEAKER_01

a big, obvious platform shift. The funny thing about mobile is that some companies missed it completely. And for some of them,

36:46

SPEAKER_01

Apple, Facebook, Amazon, and say, “Hard to see a problem for them, really, with all of this.” You can certainly see questions for all of them, and one of them may drop the ball, but it's worth remembering what happened in mobile. The internet was just a big, obvious platform shift. The funny thing about mobile is that some companies missed it completely, and for some of them, it really didn't change anything. So, for Google, it didn't change anything. For Meta, this was great. This is a way better way to do social than on PC. You've got a camera and notifications, and it's on your phone all the time with you. Amazon—what does this change? It doesn't change anything. I'm massively oversimplifying here, but the point is, now, meanwhile, Yahoo Mail fails to make the jump. There are companies that were already dying that fail to make the jump. Maybe eBay. You can argue about individual names. The point is that we went through that shift, and it didn't change anything for half the industry, half the internet industry. And so I think that you could propose a little bit of that here. It's what Steven Sinofsky at a16z, who used to run Windows, would always say: incumbents always try and make the new thing a feature. And sometimes they're right. Sometimes it's a feature.

36:47

SPEAKER_01

[SPEAKER_00] Actually, along those lines, something I wanted to get your take on: There's this thread that's been happening across a bunch of guests, which is around distribution becoming a bigger and bigger moat because, as software is easier to build, everyone's launching products. Everyone's trying to compete for attention. It's getting harder and harder. It's always been hard to get people's attention, but the noise in the market is just going up like crazy. And to me, that tells me distribution is becoming a more and more valuable skill and asset. And it also tells me incumbents are going to be a lot more successful because they already have distribution, versus a startup that's trying to break through.

36:48

SPEAKER_01

Yeah. There's a version of the Drake meme. He says, “I don't like that. I do like this.” “I don't like thin GPT wrappers. I do like harnesses.” I did spend some time talking about this in the presentation I did at the end of last year: that if the product is a commodity, then the distribution is what matters. And I wrote a thing about SAP GPT earlier this year, opening out earlier this year. How do they compete? There's an obvious comparison here that a lot of people made with web browsers. They're fundamentally a web browser. And there is a distinction here, I think, between the web browser as a product and the web browser rendering engine. And the rendering engine can be better or worse. But the browser product is just a really thin wrapper for a rendering engine. There's an input box and an output box. And what else? And what's the last innovation in browser design? Tab browsing, which was 20 years ago, 25 years ago. Every now and then, somebody tries to innovate in browser design, and it never works because you found the Platonic ideal. It's trying to innovate in smartphone design. It's a glass rectangle. There's nothing you can do there. And so, what happened, of course, is that Microsoft uses distribution to break in. Then, every now and then,

36:48

SPEAKER_01

Trying to innovate [SPEAKER_00] [SPEAKER_01] in smartphone design. [SPEAKER_00] [SPEAKER_01] It's a glass rectangle. [SPEAKER_00] [SPEAKER_01] There's nothing you can do there. [SPEAKER_00] [SPEAKER_01] And so [SPEAKER_00] [SPEAKER_01] what happened, [SPEAKER_00] [SPEAKER_01] of course, [SPEAKER_00] [SPEAKER_01] is that Microsoft [SPEAKER_00] [SPEAKER_01] uses distribution [SPEAKER_00] [SPEAKER_01] to break in. [SPEAKER_00] [SPEAKER_01] Then, [SPEAKER_00] [SPEAKER_01] every now and then, [SPEAKER_00] [SPEAKER_01] somebody tries to innovate [SPEAKER_00] [SPEAKER_01] in browser design, [SPEAKER_00] [SPEAKER_01] and it never works

37:03

SPEAKER_00

[SPEAKER_01] because [SPEAKER_01] you found the Platonic ideal. [SPEAKER_01] It's [SPEAKER_01] trying to innovate [SPEAKER_01] in smartphone design. [SPEAKER_01] It's a glass rectangle. [SPEAKER_01] There's nothing you can do there. [SPEAKER_01] And so [SPEAKER_01] what happened, [SPEAKER_01] of course, [SPEAKER_01] is that Microsoft [SPEAKER_01] uses distribution [SPEAKER_01] to break in. [SPEAKER_01] Then, [SPEAKER_01] of course, [SPEAKER_01] what also happens, [SPEAKER_01] setting aside the lawsuit, [SPEAKER_01] is that it turns out [SPEAKER_01] that winning browsers [SPEAKER_01] doesn't matter anyway [SPEAKER_01] because the value [SPEAKER_01] is further up the stack.

37:25

SPEAKER_00

[SPEAKER_01] And so [SPEAKER_01] Microsoft will use browsers [SPEAKER_01] for five, [SPEAKER_01] six years, [SPEAKER_01] and it doesn't matter. [SPEAKER_01] It doesn't get them anything. [SPEAKER_01] And so, [SPEAKER_01] clearly, [SPEAKER_01] what's happening now [SPEAKER_01] is that Google [SPEAKER_01] is using distribution [SPEAKER_01] to drive Gemini. [SPEAKER_01] And [SPEAKER_01] what's the difference [SPEAKER_01] between Gemini and Claude? [SPEAKER_01] And if you're [SPEAKER_01] using this stuff [SPEAKER_01] all day, [SPEAKER_01] then you know. [SPEAKER_01] But for a [SPEAKER_01] normal person, [SPEAKER_01] there's no difference. [SPEAKER_01] And the same thing

37:50

SPEAKER_00

[SPEAKER_01] with Meta. [SPEAKER_01] You look at survey data [SPEAKER_01] on which LLMs people use. [SPEAKER_01] Even before [SPEAKER_01] the new thing, [SPEAKER_01] the Llama thing, [SPEAKER_01] Meta was [SPEAKER_01] behind. [SPEAKER_01] It was up there [SPEAKER_01] between ChatGPT [SPEAKER_01] and Gemini, [SPEAKER_01] which, if you're in tech, [SPEAKER_01] people have completely [SPEAKER_01] written off. [SPEAKER_01] But [SPEAKER_01] they'd sprayed it [SPEAKER_01] on every surface. [SPEAKER_01] And it wasn't that bad. [SPEAKER_01] It was fine. [SPEAKER_01] So distribution [SPEAKER_01] of an adequate product, [SPEAKER_01] when the field [SPEAKER_01] is a commodity,

38:15

SPEAKER_00

[SPEAKER_01] distribution and brand [SPEAKER_01] become a big deal. [SPEAKER_01] You can see that [SPEAKER_01] in— [SPEAKER_01] you could see that [SPEAKER_01] in the [SPEAKER_01] strategy. [SPEAKER_01] OpenAI's strategy [SPEAKER_01] late last year [SPEAKER_01] was— [SPEAKER_01] people called it [SPEAKER_01] “everything everywhere [SPEAKER_01] yesterday.” [SPEAKER_01] And so they were just [SPEAKER_01] trying everything [SPEAKER_01] to work out [SPEAKER_01] how they would get that.

38:28

SPEAKER_01

How can we get a flywheel? How can we get distribution? How can we get something that sticks? How can we get people to use something before Google and Meta and Amazon spray it everywhere and get everybody using that one? And then you've got the inertia and the power of the default, and why would you switch? Obviously, Apple is the last penny to drop here. There was this slightly weird OpenAI deal, and now there's an even weirder story that OpenAI wants to sue Apple. Good luck with that. The funny thing about the Apple deal, I think, is—just not to go off on a tangent— if you go back and watch the WWDC from 2024, the whole second half of it is Apple Intelligence. That was

39:13

SPEAKER_01

the most compelling vision of a personal AI assistant. OpenAI deal, and now there's an even weirder story that OpenAI wants to sue Apple. Good luck with that. The funny thing about the Apple deal, I think, is—just not to go off on a tangent— if you go back and watch the WWDC from 2024, the whole second half of it is Apple Intelligence. That was the most compelling vision of a personal AI assistant. It's still, still the most compelling vision I've seen. They then couldn't ship it, but neither could anybody else. And you watch it again, and you're: Okay, so you want tool-using, agentic, on-device AI with no prompt injection and no hallucinations,

39:46

SPEAKER_01

and a completely standardized API system across 10,000 apps with intents that all work perfectly. And, well, that sounds good to me, but I'm not surprised they couldn't ship it. But nobody else has shipped that. But that vision was great. I really want to see what happens at WWDC in a month. Do they actually ship that now? [SPEAKER_00] Powered by Gemini. [SPEAKER_00] But that's also another point. There's going to be the AI intelligence, whatever we call it, Gemini intelligence on Android. And then there's going to be Apple Intelligence on iOS, which is powered by Gemini, but it's not going to be the same set of products. The model's just— Do they actually ship that now?

40:27

SPEAKER_01

[SPEAKER_00] Powered by Gemini. [SPEAKER_00] But that's also another point: Okay, there's going to be the AI intelligence, whatever we call it, Gemini intelligence on Android. And then there's going to be Apple Intelligence on iOS, which is powered by Gemini, but it's not going to be the same set of products.

40:37

SPEAKER_00

[SPEAKER_01] The model's just [SPEAKER_01] the dumb thing underneath— [SPEAKER_01] the funny way of putting it— [SPEAKER_01] the dumb thing underneath [SPEAKER_01] that powers the feature. [SPEAKER_01] The model's the commodity [SPEAKER_01] that powers different decisions [SPEAKER_01] about what the feature should be [SPEAKER_01] and what the distribution should be. [SPEAKER_01] And, [SPEAKER_01] in that situation, [SPEAKER_01] of course, [SPEAKER_01] Apple's got [SPEAKER_01] a billion devices [SPEAKER_01] that can run this on edge. [SPEAKER_01] And, [SPEAKER_01] and Google has this wonderful [SPEAKER_01] marketing slogan: [SPEAKER_01] coming soon

40:56

SPEAKER_01

[SPEAKER_00] [SPEAKER_01] to our most powerful devices. [SPEAKER_00] [SPEAKER_01] Meaning [SPEAKER_00] [SPEAKER_01] it won't work on most Androids. [SPEAKER_00] [SPEAKER_01] So, again, distribution questions. [SPEAKER_00] Interesting. [SPEAKER_00] Google I/O's next week, [SPEAKER_00] so we'll see what they launch. [SPEAKER_00] Oh, no, [SPEAKER_00] they launched— they launched, they launched Android. It just shows how— [SPEAKER_00] how, [SPEAKER_00] how, [SPEAKER_00] how, [SPEAKER_00] how modern it is today. Well, no, they launched it last week. Which is— it just illustrates how much we've— we've stopped paying attention to Android and iPhone, and iPhone.

41:21

SPEAKER_01

Google did have a whole big thing last week. They've got— they're replacing Chromebooks with Google Books, and they've got a new Android intelligence powered by Gemini that will roll out to the five people who bought a Pixel phone. [SPEAKER_00] So, you don't work for Google. [SPEAKER_00] Yeah. [SPEAKER_00] I want to go in a slightly different direction. [SPEAKER_00] Something that I'm curious if you're following is the anti-AI sentiment that feels like it's growing. [SPEAKER_00] It feels like, if you've seen these surveys, [SPEAKER_00] AI is less popular than ICE. [SPEAKER_00] People are trying to stop data centers from being built.

41:31

SPEAKER_01

[SPEAKER_00] I think Eric Schmidt just did a commencement speech, and people were booing him every time he mentioned AI. [SPEAKER_00] Where do you think— [SPEAKER_00] what do you think is going on? [SPEAKER_00] Where do you think this— [SPEAKER_00] how does this go over time? It's interesting. And it's a big, fuzzy mass of different stuff. I think there is something tangible: my electricity bill went up, which actually applies in a very small number of places, objectively, but it did. And this is a question. The water thing is weird because it's just completely fake. And I should call it— [SPEAKER_00] What do you think is going on? [SPEAKER_00] Where do you think this—

41:45

SPEAKER_01

[SPEAKER_00] How does this go over time? It's interesting. And it's a big, fuzzy mass of different stuff. I think there is something tangible: my electricity bill went up, which actually applies in a very small number of places, objectively, but it did. And this is a question. The water thing is weird because it's just completely fake. And I should call it— explain what I mean here. Data centers use water for cooling. It's mostly closed-loop. But the number of data centers relative to the total amount of water use in the USA is tiny. I actually went and dug into this at the Livermore Lab. They did a study at the end of 2024,

42:08

SPEAKER_01

where they estimated U.S. data center water consumption. And it came out at about 0.017% of U.S. water consumption. Now, obviously, if you live in a small town and you've got one well, and they capped the well and gave all the water to the data center, then you're really pissed off. But that's— that's a planning problem. That's not a data center problem. In generality, yes, this is— data centers are, what, 5% of U.S. energy and might grow 1% a year for the next five years. One percentage point a year. But the water stuff is just nonsense. And then you get into more tangible questions: well, what is happening with this? Is it taking jobs away?

42:28

SPEAKER_01

You can watch a bunch of three-hour podcasts of economists talking to each other. And the main answer is we really don't know yet. There's a bunch of charts that say yes, and a bunch of charts that say no. And clearly there's a slowdown in employment among 18- to 24-year-olds, but that seems to be the same for people who do and don't have degrees. And the same for people in fields that look exposed to AI and fields that don't look exposed to AI. So there's a lot of econometric argument about this. And there's a broader point here. In fact, it's a different point here: we have very little data on what's going on in AI from anyone. The model labs don't tell us anything.

42:43

SPEAKER_01

They don't give us any meaningful usage information. They give us these weird studies of people— how many people use this for this and that. They don't give us a daily active use number. We do not have a daily active user number for ChatGPT. It's crazy. And all the data comes from academic economists trying to back stuff out of BLS surveys, or consultancies and marketing agencies spending a whole bunch of money to survey 20,000 people and saying, "What are you doing with this stuff?" The model labs don't tell us anything. They don't give us any meaningful usage information.

42:56

SPEAKER_00

[SPEAKER_01] They give us these weird studies of people: [SPEAKER_01] how many people use this for this and that. [SPEAKER_01] They don't give us a daily active use number. [SPEAKER_01] We do not have a daily active user number for ChatGPT. [SPEAKER_01] It's crazy. [SPEAKER_01] And all the data comes from academic economists trying to back stuff out of BLS surveys. [SPEAKER_01] Or consultancies and marketing agencies spending a whole bunch of money to survey 20,000 people and saying, [SPEAKER_01] "What are you doing with this stuff?" [SPEAKER_01] We don't have good data on what's going on and how many people are really using this.

43:06

SPEAKER_00

[SPEAKER_01] But to the employment question, [SPEAKER_01] hence, [SPEAKER_01] there's a lot of people looking through all the stuff that the U.S. Census collects and trying to work out, [SPEAKER_01] well, [SPEAKER_01] where can we see this? [SPEAKER_01] Can we see productivity? [SPEAKER_01] What can we see? [SPEAKER_01] And the answer right now, [SPEAKER_01] I think, is: [SPEAKER_01] there's no clear consensus that we're seeing an impact on jobs. [SPEAKER_01] But, of course, [SPEAKER_01] politically, [SPEAKER_01] that doesn't matter [SPEAKER_01] if you're a student and you can't get a job. [SPEAKER_01] And that clearly is an issue.

43:28

SPEAKER_00

[SPEAKER_01] Whether it's because of AI or because of Trump and tariffs, [SPEAKER_01] it's a different question. [SPEAKER_01] Then you get niche things: [SPEAKER_01] people who draw book covers for young adult romance novels are very upset that now you can get a picture of a naked woman on the back of a dragon flying through a volcano without paying them. [SPEAKER_01] So, [SPEAKER_01] I'm sorry, [SPEAKER_01] I'm being deliberately unkind,

43:35

SPEAKER_01

[SPEAKER_00] [SPEAKER_01] but there's a little— [SPEAKER_00] [SPEAKER_01] there are, [SPEAKER_00] [SPEAKER_01] there are [SPEAKER_00] [SPEAKER_01] people, [SPEAKER_00] [SPEAKER_01] particularly novelists, [SPEAKER_00] [SPEAKER_01] people who write eBooks. [SPEAKER_00] [SPEAKER_01] There's a huge culture war over whether it's okay to use AI. [SPEAKER_00] [SPEAKER_01] There's this whole AI slop question. [SPEAKER_00] [SPEAKER_01] And [SPEAKER_00] [SPEAKER_01] if you saw the number that 30–40% of your podcasts are generated by AI. [SPEAKER_00] [SPEAKER_01] So there's a lot of— [SPEAKER_00] [SPEAKER_01] there are big, fuzzy, [SPEAKER_00] [SPEAKER_01] massive questions.

43:47

SPEAKER_01

[SPEAKER_00] [SPEAKER_01] Some of this, [SPEAKER_00] [SPEAKER_01] I think, is a little bit like the backlash we had around social, [SPEAKER_00] [SPEAKER_01] but much more compressed. [SPEAKER_00] [SPEAKER_01] And like social, [SPEAKER_00] [SPEAKER_01] some of the backlash around social was true. And some of it was true. And some of it wasn't. [SPEAKER_00] [SPEAKER_01] So there's a lot of— [SPEAKER_00] [SPEAKER_01] there are big, fuzzy, [SPEAKER_00] [SPEAKER_01] massive questions. [SPEAKER_00] [SPEAKER_01] Some of this, [SPEAKER_00] [SPEAKER_01] I think, is a little bit like the backlash we had around social, [SPEAKER_00] [SPEAKER_01] but much more compressed.

43:58

SPEAKER_01

[SPEAKER_00] [SPEAKER_01] And like social, [SPEAKER_00] [SPEAKER_01] some of the backlash around social was true. And some of it was true. And some of it wasn't, always exemplified in the whole "Facebook sells your data" thing, which is just, A, not true, and, B, the people who believe it are absolutely adamant that, of course, it's true. And you're— you're obviously a lunatic for suggesting otherwise. It's the line from Jonathan Swift that you can't reason somebody out of an idea that they weren't reasoned into. So you get this wide— it was a long way out of answering the question, but you get this wide spread of ideas, just as you did with social.

44:17

SPEAKER_01

There's 20 different things, some of which are really real and some of which are really not real, and a lot of which are a fuzzy mess in the middle. All of which means that, meanwhile, you've got Trump saying he wants a new executive order on dangerous models, which I actually don't think is the thing that drives the backlash. They're worrying about bio or cyber. I don't feel that's a Main Street America conversation. And that's the thing that got Trump interested in this stuff again. [SPEAKER_00] Let me go in a tangential direction. [SPEAKER_00] Something that I like to ask [SPEAKER_00] folks who have kids who come on the podcast,

44:29

SPEAKER_01

[SPEAKER_00] especially people who are thinking so deeply about where things are going: [SPEAKER_00] Knowing what you know about where the world is heading, [SPEAKER_00] what AI is going to do to the future, [SPEAKER_00] how are you changing the way you raise your kids? [SPEAKER_00] What are you teaching them differently, [SPEAKER_00] potentially, that might help them in the future? I don't know. I think there's a curve here in that if you've got kids who are going onto the job market in the next year or two, then everything is up in the air, and no one knows how this is going to work. If you've got kids who are going onto the job market in five years, then who knows?

44:38

SPEAKER_01

But stuff will have settled down a lot by then, in probably unpredictable ways. So I could be a lot more worried if I had a 21-year-old. I don't. I've got a kid in his early teens. So it's a different— those, those questions vary. Then you've got a lot of the questions that were the same before ChatGPT around the collapse of gatekeepers, the— no, should you really believe what that influencer on TikTok says? And where exactly are you getting your understanding of what's going on in Israel? And all of those kinds of social media, internet, media consumption kinds of questions. I don't know. There are people who are super, super intentional about

44:54

SPEAKER_01

every minute of their child's life. I'm not. I recall the George Carlin line that anyone who drives faster than you is a maniac, and anyone who drives slower is an idiot. And that certainly applies to parenting. And all of those kinds of social media, internet, media consumption kinds of questions. I don't know. There are people who are super, super intentional about every minute of their child's life. I'm not. I recall the George Carlin line that anyone who drives faster than you is a maniac, and anyone who drives slower is an idiot. And that certainly applies to parenting. And so I— everybody thinks they're somewhere in the middle, but I don't have

45:12

SPEAKER_01

a deeply systematic, widespread, and coherent plan for: This is what my child is going to be doing in three, six, 12, 18 months' time. I'd settle for him not breaking his Chromebook again. [SPEAKER_00] The church's general vibe is: It's going to be okay. [SPEAKER_00] Guys, [SPEAKER_00] it's going to be okay. Yeah. I don't know if you—I think if you— maybe this is because I'm British, and we haven't had political violence in 500 years. And I think maybe if I came from Iran, I'd have a different attitude to being calm about the future. I think there's a layer of: Yes, this will change a bunch of stuff, and we'll need to worry about it. But that's a constant.

45:30

SPEAKER_01

We've always had that. I remember, in the whole wave of the panic around social media, I dug up— so there were a whole bunch of books in the late seventies about databases. There was a whole panic about databases. And again, half of it was true. If everybody's police records and a restaurant— if all police records and all government records are online, then that's different. If you think about, for example, the deep nudes— deepfake nudes issue, for example, there's a dumb reaction to this, which is to say, “Haven't you heard of Photoshop?” Which is true, There was a whole panic about databases. And again, half of it was true.

45:48

SPEAKER_01

If everybody's police records and a restaurant— if all police records and all government records are online, then that's different. If you think about, for example, the deep nudes— deepfake nudes issue, for example, there's a dumb reaction to this, which is to say, “Haven't you heard of Photoshop?” Which is true, but a 15-year-old kid couldn't use Photoshop to make hardcore pornographic nudes of every girl in their high school and send them to the whole school in one afternoon. And turn them into video. Exactly. Even—well, yeah, even more. And now they can. So that is different. It's the challenge of social— the thing people would say in the nineties is, “It's great.

46:07

SPEAKER_01

You can be the only gay kid in your village, and you can find other gay people, and you can find your tribe.” And guess what? It turned out you could also be the only Nazi in your village, or the only pedophile in your village, or somebody who wanted to look at child porn. And yeah, now you can find the other people who like looking at child porn, and they'll tell you it's great. So, oops, we connected everybody. And unfortunately, that meant we connected all the bad people and all of our own worst instincts and every problem in society. And so that will happen again with AI. Deepfake nudes are the obvious thing we can see now.

46:20

SPEAKER_01

There will be a whole bunch more of this stuff. But there's also something a technical audience should know about. Do you know about the Post Office scandal in the UK? Nope. Okay. So, sidebar here. So in the UK, post offices are mostly franchises run by small businesspeople. So they're run by pharmacies, classically. Very often Indian immigrants, second-generation Indian people. And so the Post Office, 15 years ago, rolled out a new point-of-sale computer system. So they have a separate counter in the back. That's the post office. And so the Post Office rolled out this new computer system built by Fujitsu that had a bunch of bugs in it that showed shortfalls in cash.

46:33

SPEAKER_01

And the Post Office looks at this and says, “Aha, we knew these people were stealing from us.” Hundreds of people go to prison. A bunch of suicides. A bunch of bankruptcies. People lose their homes. Meanwhile, people from the Post Office and people from Fujitsu are going to court and swearing there are no bugs in the system and nobody else has had this problem. This is 1970s technology. And that's really the point: Every wave of technology comes with ways that you can ruin people's lives, either deliberately or by accident. This is the whole thing of Chinese mass surveillance: It's deliberate. This is—maybe people should go to prison. Maybe not.

46:50

SPEAKER_01

But we have this with every technology. We have a bunch of ways that you can ruin people's lives. And you have to be conscious of that and also not panic about it. [SPEAKER_00] So maybe following that thread and coming back to the kids thing and the jobs thing: [SPEAKER_00] Is there a job you are steering your kid away from? [SPEAKER_00] And is there a job you think you want to steer them towards? I don't know about that. It's probably a little bit early yet. He's not quite at the “I want to be a fireman” stage. [SPEAKER_00] That might be a great job. Yeah. And certainly— And you have to be conscious of that and also not panic about it.

46:57

SPEAKER_01

[SPEAKER_00] So maybe following that thread and coming back to the kids thing and the jobs thing: [SPEAKER_00] Is there [SPEAKER_00] a job you are steering your kid away from? [SPEAKER_00] And is there a job you think you want to steer them towards? I don't know about that.

47:02

SPEAKER_00

[SPEAKER_01] It's probably a little bit early yet. [SPEAKER_01] He's not quite at the

47:04

SPEAKER_01

“I want to be a fireman” stage. [SPEAKER_00] That might be a great job. Yeah. And certainly, if I look at my career, I started as an equity analyst, and then I went and worked in industry, and then I was a consultant. The days when you knew what your career was going to be. However, there were certainly some people where you want to be an architect, you want to be a software engineer, you want to be X or Y. I don't know. I think the only, the only thinking I have here is that you slowly work out there's a bunch of skills that you have, and there's a bunch of jobs that those skills make you good at.

47:19

SPEAKER_01

And then there's a bunch of stuff that people will pay you for, and you want to get at least two of those, and preferably all three. [SPEAKER_00] Okay. [SPEAKER_00] So zooming out a little bit, [SPEAKER_00] let me ask you a meta question. [SPEAKER_00] What's a question about AI that you think nobody's asking yet, or not enough people are asking, that we should be asking ourselves? Sure. Versus what are people actually hiring you for? Is that a useful way of thinking about this? And clearly there are going to be some jobs where, no, that is just a task, and that job gets automated away. But there's a bunch where that isn't the question.

47:29

SPEAKER_01

The way I actually pulled that together at the end of the deck was a chart of global recorded music revenue, which, you may know, is a U-shaped curve, more or less. [SPEAKER_00] What's a question about AI that you think nobody's asking yet, or not enough people are asking, that we should be asking ourselves? Sure. Versus what are people actually hiring you for? Is that a useful way of thinking about this? And clearly there are going to be some jobs where, no, that is just a task, and that job gets automated away. But there's a bunch where that isn't the question. The way I actually pulled that together at the end of the deck was a chart of global recorded music revenue,

47:40

SPEAKER_01

which is a U-shaped curve, more or less. And so it dropped by about half from 2000 to 2015 or so. And since then, it has come back to about 75% of the peak, adjusted for inflation.

47:42

SPEAKER_00

[SPEAKER_01] And the way that I look at this is to say— [SPEAKER_01] and that's driven by streaming— [SPEAKER_01] I looked at this and said, [SPEAKER_01] the first half of this chart is saying, [SPEAKER_01] what happens if I don't have to pay $15 to get a CD to get that track?

47:46

SPEAKER_01

And the second half of the chart is saying, what happens if $15 a month gives you all the music that there is? [SPEAKER_00] [SPEAKER_01] So it's a completely different question.

47:49

SPEAKER_00

[SPEAKER_01] And you could— [SPEAKER_01] that's the way that you could look at Uber or the way you could look at Airbnb, [SPEAKER_01] all these kinds of companies—is that, to begin with, [SPEAKER_01] you do the old thing,

47:51

SPEAKER_01

but more. With any new technology, you do the old thing, but more of it in the new place. So you put Flickr on mobile, you print out your emails, and then you make new things that are only possible with the new thing. And then maybe you go a bit further and completely redefine the question, and you make something that isn't that at all. Spotify is not an online music store. It's something else. And right now, those questions—

48:09

SPEAKER_00

[SPEAKER_01] you only even know what the question is after it's been asked and you've

48:10

SPEAKER_01

built a billion-dollar thing that lots of people use.

48:11

SPEAKER_00

[SPEAKER_01] Because [SPEAKER_01] obviously Spotify looked crazy, and Uber looked crazy, and Airbnb looked crazy. [SPEAKER_01] But that's, [SPEAKER_01] I think, the way to get at what this means: you have to get past [SPEAKER_01] "We do the old stuff, [SPEAKER_01] but more." [SPEAKER_01] And you have to get to, [SPEAKER_01] "What do you do that's different [SPEAKER_01] because of this? [SPEAKER_01] What does this change? [SPEAKER_01] What wasn't possible before? [SPEAKER_01] What gets unlocked?" [SPEAKER_01] As opposed to just doing the old thing,

48:37

SPEAKER_01

but more of it. [SPEAKER_00] Yeah. [SPEAKER_00] Just to support this general theme you have: [SPEAKER_00] we don't know what is going to happen. [SPEAKER_00] This is unprecedented. [SPEAKER_00] If you, [SPEAKER_00] if you were to zoom out [SPEAKER_00] a few years ago, [SPEAKER_00] maybe three years ago, [SPEAKER_00] four years ago, [SPEAKER_00] the last profession you'd think would be automated is engineering and coding. [SPEAKER_00] That feels like the hardest thing. [SPEAKER_00] We're going to need people to build these things. [SPEAKER_00] Now it's the most transformed role of any role.

49:08

SPEAKER_01

[SPEAKER_00] You went from writing all your code to 0% of your code being written by AI. [SPEAKER_00] It's almost as if you didn't realize— [SPEAKER_00] [SPEAKER_01] you didn't realize it was boring manual labor that could be automated. [SPEAKER_00] [SPEAKER_01] You thought it was something else. [SPEAKER_00] [SPEAKER_01] It's funny. [SPEAKER_00] [SPEAKER_01] I was looking at this whole— there's a US government-owned data set called O*NET or something like that, which tries to analyze every single job. And then people try and score it. And they try and say, this profession is X or Y percent exposed to AI, and AI can do Z percent of it today.

49:35

SPEAKER_01

I think this is just the most ridiculous bunch of deluded horseshit. And there are two reasons for this. The first reason is that this is, ironically, the logical systems problem. The expert systems problem. [SPEAKER_00] [SPEAKER_01] The problem with expert systems is, for anyone who doesn't know, you try to recognize a picture of a cat. [SPEAKER_00] [SPEAKER_01] And so you start building up logical steps. [SPEAKER_00] [SPEAKER_01] And they try and say, [SPEAKER_00] [SPEAKER_01] this profession is X or Y percent exposed to AI, and AI can do Z percent of [SPEAKER_00] [SPEAKER_01] it today.

49:53

SPEAKER_01

[SPEAKER_00] [SPEAKER_01] I think this is just the most ridiculous bunch of deluded horseshit. [SPEAKER_00] [SPEAKER_01] And there are two reasons for this. [SPEAKER_00] [SPEAKER_01] The first reason is that, [SPEAKER_00] [SPEAKER_01] ironically, [SPEAKER_00] [SPEAKER_01] this is the logical systems problem. [SPEAKER_00] [SPEAKER_01] The expert systems problem. [SPEAKER_00] [SPEAKER_01] The problem with expert systems is, for anyone who doesn't know, [SPEAKER_00] [SPEAKER_01] you try to recognize a picture of a cat. [SPEAKER_00] [SPEAKER_01] And so you start building up logical steps. So you'd make an edge detector, and then you make a third detector, and you make

50:17

SPEAKER_01

an eye detector, and you make an ear detector. And 15 years later, you've got 700 steps, and it doesn't work. And this is what happens when you try and look at a profession and break it down by which bits can be automated and which can't. You can't describe a profession like that. Or, at any rate, we can't. You can't look at a senior partner at a law firm and say, 17% of their work could be automated. This is horseshit. You can't do that. I think the other side of the fallacy, though, is to talk about taxi drivers. So, You've got 700 steps, and it doesn't work. And this is what happens when you try to look at a profession and

50:51

SPEAKER_01

break it down by which bits can be automated and which can't. You can't describe a profession like that. Or, at any rate, we can't. You can't look at a senior partner at a law firm and say, 17% of their work could be automated. This is horseshit. You can't do that. I think the other side of the fallacy, though, is to talk about taxi drivers. So, if we'd been having this conversation in 1997, it's the Uber test. Imagine we're in 1997. What will be crushed by the internet? Newspapers will be fine. They'll just— because they'll save money on the printing bills. This is a joke, but people said the internet would be great for newspapers. Their printing bills will go down. Yes,

51:27

SPEAKER_01

but no. But the other side is, obviously, taxi drivers— you couldn't automate that with the internet. It's got nothing to do with the internet. Maybe you'd have internet booking, but that's not going to change anything. And, of course, it completely changes the whole thing. And so the— the example I saw the other day was things that won't be affected by AI: personal trainers. Okay. So I take my iPhone and balance it on the metal piece with the camera pointed at me. And I ask an AI to build me a training routine, watch me, and tell me if I'm doing it right. Why do I need a personal trainer? Now, that might be complete nonsense. But that's how these things work.

51:57

SPEAKER_01

The stuff that you don't think is— you can't necessarily predict which things are going to be exposed. Or, a lot of the big companies are things that didn't look like they would work and didn't look like they were exposed. The other side of this, of course, is—this is one of the charts at the end of my presentation—is comparing Uber and Airbnb, because this is the cliché from Marc Andreessen that Uber doesn't sell software to taxi companies. Airbnb doesn't sell software to hotels. Okay. Now let's go and look at the market impact. There are a whole bunch of cities where Uber demolished the taxi business and made it much bigger as well.

52:15

SPEAKER_01

The TAM became much bigger, and everyone switched. Airbnb's impact on hotels, if you actually go and look at the numbers, is pretty marginal. They carved out this whole other business, and maybe they slowed down the growth of hotels a bit. But my wife flies to Milwaukee next week. She's going to land at eight o'clock at night. She wants to go to a hotel. She wants to have room service. She needs a bathroom, a bath. She needs a gym at six in the morning. And then, at seven in the morning, she's going to drive to the client site. She's not going to stay in an Airbnb. Absolutely zero chance she's going to stay in an Airbnb. And half of the hotel business is travel—

52:45

SPEAKER_01

is business travel. And as soon as you actually get into anything, then it gets complicated. I remember somebody on social media said, “The problem with Benedict is his answer to everything is, ‘It depends.’” Yeah, it does. It depends. So there were— it's back to my 1997 point. You can say some of this, [SPEAKER_00] but you have to have that humility. [SPEAKER_00] Yeah.

53:12

SPEAKER_00

[SPEAKER_01] [SPEAKER_00] I'm coming back to this phrase you use: [SPEAKER_01] [SPEAKER_00] “Presume radical uncertainty” is a nice core thesis here. [SPEAKER_01] [SPEAKER_00] So, knowing all this, [SPEAKER_01] [SPEAKER_00] it's hard to tell. [SPEAKER_01] [SPEAKER_00] We don't know exactly where it's going. [SPEAKER_01] [SPEAKER_00] Things are going to change a lot, [SPEAKER_01] [SPEAKER_00] but it'll probably be okay broadly. [SPEAKER_01] [SPEAKER_00] A lot of people listening are pretty worried about their jobs and their careers [SPEAKER_01] [SPEAKER_00] and how much the world changes.

53:33

SPEAKER_01

[SPEAKER_00] What would be a couple of things you recommend people do, knowing what you know, to be more [SPEAKER_00] successful in this future? Well, I should just wind back on what you just said. As Keynes tells us, in the long run, we're all dead. So, it's all— on average, on average, nobody died in World War One. Great. But if— if you're— [SPEAKER_00] Successful in this future? Well, I should just wind back on what you just said. As Keynes tells us, in the long run, we're all dead. So, it's all— on average, on average, nobody died in World War One. Great. But if— if you're— if you're a 19-year-old in 1914, you've got a one-in-three chance of not coming back. So, yes,

54:36

SPEAKER_01

clearly there's a bunch of professions where this is a major question. And particularly if you're an associate or would have been thinking about being an associate, this is a major question. And it's very unclear how those professions are going to play out. It's very unclear what happens to the pyramid structure of professional services. The answer— the only answer I think one can have is— If you're— if you're a 19-year-old in 1914, you've got a one-in-three chance of not coming back. So, yes, clearly there's a bunch of professions where this is a major question. And particularly if you're an associate or would have been thinking about being an associate,

54:55

SPEAKER_01

this is a major question. And it's very unclear how those professions are going to play out. It's very unclear what happens to the pyramid structure of professional services. The answer— the only answer I think one can have is:

55:09

SPEAKER_00

[SPEAKER_01] Don't stick your head in the sand and say, [SPEAKER_01] "I hate all of this stuff," [SPEAKER_01] because that gives you a great feeling of moral superiority.

55:15

SPEAKER_01

And you can go on Bluesky and shout at everybody, shout at each other about how evil AI is. Great, I'm happy for you. But that's not going to help. What helps is you diving into this, completely submerging yourself in it, and coming out understanding what you can do with it, how this changes things, how you can be a great hire. [SPEAKER_00] [SPEAKER_01] And that may still not help. [SPEAKER_00] [SPEAKER_01] But [SPEAKER_00] [SPEAKER_01] if you're going into a law firm and they're saying, [SPEAKER_00] [SPEAKER_01] "Well, [SPEAKER_00] [SPEAKER_01] we hired a hundred associates last year, and this year we're only going to hire 50,"

55:39

SPEAKER_01

[SPEAKER_00] [SPEAKER_01] going to the interview and saying, [SPEAKER_00] [SPEAKER_01] "Well, [SPEAKER_00] [SPEAKER_01] I think AI is bullshit, and I'm never going to use it," is probably not the right [SPEAKER_00] [SPEAKER_01] move. That, that, that may not be particularly comforting, but I don't think there's an alternative. You have to dive into this and absorb it and internalize it and think about what it means, just as you and I did with mobile and with the internet. [SPEAKER_00] I think that is actually very actionable and [SPEAKER_00] very consistent advice on the podcast. Just— [SPEAKER_00] just do stuff, [SPEAKER_00] build it,

56:10

SPEAKER_01

[SPEAKER_00] don't just sit around and pontificate and be [SPEAKER_00] pissed at what's happening. [SPEAKER_00] To [SPEAKER_00] close this out, [SPEAKER_00] I'm going to take us to AI Corner, [SPEAKER_00] a recurring corner of the podcast. [SPEAKER_00] And the question to you is just: What's one way you used AI and use AI in your [SPEAKER_00] work or life [SPEAKER_00] that [SPEAKER_00] is really interesting, [SPEAKER_00] something that other people might, [SPEAKER_00] might [SPEAKER_00] be inspired by? I don't know. I— I struggle with this question because I'm the lawyer looking at ChatGPT. So, the stuff that I would do, that I would automate, are precise

56:38

SPEAKER_01

information retrieval tasks, which is precisely the thing that this is worst at. And that's not a criticism. It's just an observation. The stuff that I would want a machine to do for me is the stuff that AI can't do for me very, very, very well at the moment. I use it for proofreading. I use it for images. I used it while redecorating my apartment. That worked fantastically well. "Here's a picture of this room repainted with this light and this table and this rug. No, change the color of the rug." There are parts of stuff where it works. But a couple of years ago, somebody said AI is good at stuff that computers are bad at and bad at stuff that computers are good at.

57:12

SPEAKER_01

And that's— [SPEAKER_00] [SPEAKER_01] I— [SPEAKER_00] [SPEAKER_01] I— [SPEAKER_00] [SPEAKER_01] I struggle to find many, many examples of those where I need it. But then, I have a unique, weird job. I sit at my desk all day, trying to synthesize a whole bunch of other stuff into a whole bunch of new ideas. That's not a particularly common way for people to spend their time. I struggle to find AI use cases. I am the accountant looking at the spreadsheet and thinking, "Well, that's very clever. And this is clearly going to completely transform everything. But I actually don't make spreadsheets every day." [SPEAKER_00] I went to a stand-up comedy show with

57:47

SPEAKER_01

[SPEAKER_00] Pete Holmes. [SPEAKER_00] I don't know if you know him. [SPEAKER_00] And he made this joke that [SPEAKER_00] we want AI to clean the poop off the street and do all these hard things [SPEAKER_00] that nobody wants to do, [SPEAKER_00] but instead it's, [SPEAKER_00] "Oh, [SPEAKER_00] let me help you write. [SPEAKER_00] Let me help you create imagery." [SPEAKER_00] It's this bohemian. [SPEAKER_00] It's, [SPEAKER_00] "No." [SPEAKER_00] I went to a stand-up comedy show with Pete Holmes. [SPEAKER_00] I don't know if you know him.

58:22

SPEAKER_01

[SPEAKER_00] And he made this joke that we want AI to clean the poop off the street and do all these hard things that nobody wants to do, but instead it's: [SPEAKER_00] "Oh, let me help you write.

58:27

SPEAKER_00

[SPEAKER_01] [SPEAKER_00] Let me help you create imagery." [SPEAKER_01] [SPEAKER_00] It's this bohemian. [SPEAKER_01] [SPEAKER_00] It's: [SPEAKER_01] [SPEAKER_00] “No.”

58:39

SPEAKER_01

[SPEAKER_00] I went to a stand-up comedy show with Pete Holmes. [SPEAKER_00] I don't know if you know him. [SPEAKER_00] And he made this joke that we want AI to clean the poop off the street and do all these hard things that nobody wants to do, but instead it's: [SPEAKER_00] “Oh, let me help you write.

58:47

SPEAKER_00

[SPEAKER_01] [SPEAKER_00] Let me help you create imagery.”

58:49

SPEAKER_01

[SPEAKER_00] It's this bohemian. [SPEAKER_00] It's: [SPEAKER_00] “No, I don't want to do all these ugly things. [SPEAKER_00] I want to be creative. [SPEAKER_00] Make art.” Yeah. Well, there's, there's, there's variations of all of this. It's: “I don't want the AI to do the stuff I do for fun. I want it to do the boring stuff that I don't do for fun.” Yeah. And finding that mesh—joking apart, this is going to come back to my chatbot, chatbot point—that the chatbot is a blank screen and a jagged edge. What am I supposed to do, and what will work? And that's a big problem. And the solution to that problem is to wrap it in use cases.

59:03

SPEAKER_01

Part of it is also that AI just disappears. So most of what I write now, I dictate. I dictate as a voice memo, and that's automatically transcribed. Is that still AI, or is that just voice recognition? Probably an LLM. There's probably an LLM in there. Okay. So maybe that's AI.

59:21

SPEAKER_00

[SPEAKER_01] Well, okay. [SPEAKER_01] So, so what? [SPEAKER_01] [SPEAKER_00] At a certain point, it's just automation. [SPEAKER_01] [SPEAKER_00] What do you use for that?

1:00:02

SPEAKER_01

[SPEAKER_00] For voice, voice transcription? So I actually find Apple Notes, the app or the one built into the iPhone, works fine. I'm conscious that people want others, but I dictate it, though. It is. It worked. So I'm, I'm happy with that. [SPEAKER_00] All right. [SPEAKER_00] Final question before we get to our very exciting lightning round. [SPEAKER_00] Is there anything else that you wanted to share? [SPEAKER_00] Anything else you want to leave listeners with? No, I think I've, I've, I've monologued plenty, and I've gone through a bunch of stuff in the deck. Go read the deck and sign up for my newsletter.

1:00:37

SPEAKER_01

And then you will get many more mags of brilliant Benedict Evans wisdom, some of which may even be useful. Someone unsubscribed from my newsletter, and they said: “You didn't, you didn't give me any actionable stock ideas.” And I'm: “Well, on one level, that's completely true. On the other level, maybe not.” [SPEAKER_00] Well, with that, Benedict, we've reached our very exciting lightning round. [SPEAKER_00] I've got five questions for you. [SPEAKER_00] Are you ready? [SPEAKER_00] Sure. [SPEAKER_00] First question. [SPEAKER_00] What are two or three books that you find yourself recommending most to other people?

1:01:05

SPEAKER_01

That's a tough one for me because I just read an enormous number of books, and then I can't remember which ones I've read. I, I, I sometimes joke that the classic British comedy from the late 19th century called Three Men in a Boat is my I Ching. We're having trouble hanging a picture. Well, there's a section about that. We're having trouble doing this. Oh, well, there's a success story about that. All of which are hilarious. So Three Men in a Boat is my I Ching.

1:01:14

SPEAKER_01

There's a book by, I think, William Cronon about the economic history of Chicago, which is fascinating and actually very relevant to technology because it's talking about standardization and packetization and logistics and channel conflict and network dynamics and network neutrality. So when the meatpackers of Chicago reach the point that it's cheaper to ship a cow from New York to Chicago, kill it, pack it, and then ship it back to New York than to kill it in New York. And the pricing of refrigerator cars. And it's exactly like reading about broadband. It's all the same kind of—it's those kinds of business issues, which is fascinating. What else have I read?

1:01:26

SPEAKER_01

I don't know. Read books. Read different books. Generally, read books for grown-ups. Please read something other than Lord of the Rings. If you're going to name another company, it's: “I saw this, and what was the latest Peter Thiel company?” I was: “Read another book.” [SPEAKER_00] [SPEAKER_01] Everything is named after a character from this one book. [SPEAKER_00] [SPEAKER_01] There is more than one book in the world. [SPEAKER_00] [SPEAKER_01] It's those kinds of business issues, [SPEAKER_00] [SPEAKER_01] which is fascinating. What else have I read? I don't know. Read books. Read different books. [SPEAKER_00] [SPEAKER_01] Generally, read books for grown-ups.

1:01:58

SPEAKER_01

Please read something other than Lord of the Rings. If you're going to name another company and it's:

1:02:02

SPEAKER_00

[SPEAKER_01] “I saw this, and what was the latest Peter Thiel company?” [SPEAKER_01] I was: [SPEAKER_01] “Read another book.” [SPEAKER_01] Everything is named after a character from this one book. [SPEAKER_01] There is more than one book in the world. [SPEAKER_01] There is more than one book. [SPEAKER_01] Then all about science fiction. [SPEAKER_01] Read, [SPEAKER_01] read about different things. [SPEAKER_01] Read about things you don't know about. [SPEAKER_01] [SPEAKER_00] Along those lines, [SPEAKER_01] [SPEAKER_00] do you have a favorite recent movie or TV show that you've really enjoyed? [SPEAKER_01] I don't know. [SPEAKER_01] I've dropped so badly off the

1:02:22

SPEAKER_00

[SPEAKER_01] current media treadmill. [SPEAKER_01] And I just spend most of my time watching classics. [SPEAKER_01] "Read another book."

1:02:29

SPEAKER_01

Everything is named after a character from this one book. There is more than one book in the world. There is more than one book. They're all about science fiction. Read. Read about different things. Read about things you don't know about. [SPEAKER_00] Along those lines, [SPEAKER_00] do you have a favorite recent movie or TV show that you've really enjoyed? I don't know. I've dropped so badly off the current media treadmill. And I just spend most of my time watching classics, which are always the ones that you're supposed to have seen and that all seem intimidating.

1:02:58

SPEAKER_01

And then you watch them, and you're, "Oh, that was actually really good." [SPEAKER_00] [SPEAKER_01] I watched The Seventh Seal recently, [SPEAKER_00] [SPEAKER_01] which is one of those Jake Woody Allen, [SPEAKER_00] [SPEAKER_01] terrifying, [SPEAKER_00] [SPEAKER_01] boring movies. And it was brilliant. It was really interesting. And it's only an hour. So go watch, go watch one of those movies that you are supposed to have seen or [SPEAKER_00] hadn't seen. [SPEAKER_00] Favorite recent product that you've recently discovered that [SPEAKER_00] you really love. [SPEAKER_00] It could be a gadget. [SPEAKER_00] It could be an app.

1:03:38

SPEAKER_01

I was speaking at a partner meeting for a company earlier this week. What's today? Monday? No, last week. And I met the founder of the company, who has a very famous network. The CEO of the company has a very famous name, and I admired his shoes and didn't say anything, but then went and Googled half an hour later. Yeah. Okay. I'll buy a pair of these. Do you want to share the brand, or do you want to keep it, keep it secret? Okay. We'll keep it secret. I don't know. I think it comes in, it comes in waves of new products, and you get into waves of new things. When's the last time there was a cool app? iPhone apps, that was— all that white space went.

1:04:11

SPEAKER_01

It's partly a function of product shifts, platform shifts. All the white space went for cool new apps. And now we haven't quite got— actually, this is, to the earlier point, we don't have breakout consumer AI apps yet because, I think, because of marginal cost more than anything else, you can't make it free and get 50 million users and then have a revenue model. But we don't have those breakout things yet. For consumers. [SPEAKER_00] Yeah. For consumers. No, I just— we keep getting these ads for voice recorders. Somebody's selling a business-card-sized hardware voice recorder. I'm— [SPEAKER_00] [SPEAKER_01] "But [SPEAKER_00] [SPEAKER_01] I don't get it.

1:05:03

SPEAKER_01

[SPEAKER_00] [SPEAKER_01] I've got the voice recorder on my phone." [SPEAKER_00] Yeah. [SPEAKER_00] All kinds of cool stuff coming. [SPEAKER_00] Okay. [SPEAKER_00] Two more questions. [SPEAKER_00] Do you have a favorite life motto that you find yourself coming back to often in [SPEAKER_00] work or in life? [SPEAKER_00] [SPEAKER_01] I suppose I've mentioned it earlier. [SPEAKER_00] [SPEAKER_01] Apparently, I mostly say, "It depends." [SPEAKER_00] That's going to be the title. [SPEAKER_00] [SPEAKER_01] It'll probably be okay. [SPEAKER_00] [SPEAKER_01] Yeah. [SPEAKER_00] [SPEAKER_01] Okay. [SPEAKER_00] I don't— [SPEAKER_00] that's, [SPEAKER_00] that's the vibe I get.

1:05:40

SPEAKER_01

[SPEAKER_00] I like that. [SPEAKER_00] I like that. [SPEAKER_00] It's probably going to be okay. [SPEAKER_00] Not for sure. [SPEAKER_00] Okay. [SPEAKER_00] Final question. [SPEAKER_00] I saw somewhere that you own a lot of old phones. [SPEAKER_00] Is that true? It is. Yes. As a— I kept— I was a telecoms analyst and mobile analyst, and I kept all my phones up to a point. And now they're uninteresting, but as you may remember, before the iPhone, particularly outside the USA, there was this huge creativity and expansion in what phones looked like, because everyone was innovating around a little teeny-tiny gray square.

1:06:05

SPEAKER_01

So everyone was trying to differentiate from everything else. Before it results— it's like cars, actually. It's like cars before street— before wind tunnels. Cars all looked different. And everyone's trying to innovate because you've got the same four wheels and the same engine.

1:06:18

SPEAKER_00

[SPEAKER_01] And everyone's trying to differentiate based on the shape, and [SPEAKER_01] then everything converges on one shape. [SPEAKER_01] And it's the same with phones. [SPEAKER_01] Everyone— [SPEAKER_01] everything converged on one shape. Before that, [SPEAKER_01] it was all this innovation. [SPEAKER_01] So yeah, [SPEAKER_01] I [SPEAKER_01] have [SPEAKER_01] a whole bunch of PDAs and— [SPEAKER_01] It's like cars before street, [SPEAKER_01] before wind tunnels. [SPEAKER_01] Cars all looked different.

1:06:48

SPEAKER_01

And everyone's trying to innovate because you've got the same four wheels and the same engine. And everyone's trying to differentiate based on the shape, and then everything converges on one shape. Then everything converges on one shape. And it's the same with phones. Everyone— everything converged on one shape. Before that, it was all this innovation. So, yeah, I have a whole bunch of PDAs, and it's like cars before street— before wind tunnels. Cars all looked different. And everyone's trying to innovate because you've got the same four wheels and the same engine. And everyone's trying to differentiate based on the shape, and then everything converges on one shape.

1:07:07

SPEAKER_01

And it's the same with phones. Everyone— everything converged on one shape. Before that, it was all this innovation. So, yeah, I, I have a whole bunch of PDAs and smartphones. [SPEAKER_00] And how many phones are we talking about? I don't know. [SPEAKER_00] Twenty or 30. [SPEAKER_00] Okay. [SPEAKER_00] Okay. [SPEAKER_00] It's not so crazy. [SPEAKER_00] What's the oldest one? [SPEAKER_00] What's the oldest one you've got? So, I have one of those— If you'd told me, I'd have got the box down. I have one of those Ericsson shark-fin flip phones from '98 or something, which is very— again, hardware design, visual design, trying to differentiate.

1:07:56

SPEAKER_01

I've got an i-mode phone from 2001 and a J-Phone phone from 2001 that has a camera. So, I came back from Japan in 2001, and I found it had a color screen and a camera. And I just had endless client meetings, and people just wanted to see the phone with a color screen. It was mind-blowing. It didn't work outside Japan. Actually, I plugged it in the other day. It still charges up. Clearly, I can't do anything with it. And there's a little bit of an analogy in there as well. And we thought there'd be all these different shapes and sizes. And before the iPhone, people imagined, well,

1:08:25

SPEAKER_01

some people would have a little Pocket PC, and some people would have a keyboard, and you'd have folding, or there were all these different ideas for what it would look like. And it all— we didn't realize it was all going to converge on one device. [SPEAKER_00] Benedict. [SPEAKER_00] This was amazing.

1:08:33

SPEAKER_00

[SPEAKER_01] [SPEAKER_00] I learned a ton. [SPEAKER_01] [SPEAKER_00] I feel better after this conversation. [SPEAKER_01] [SPEAKER_00] Two final questions. [SPEAKER_01] [SPEAKER_00] Where can folks find you online? [SPEAKER_01] [SPEAKER_00] Where do they find this presentation? [SPEAKER_01] [SPEAKER_00] And how can listeners be useful to you? [SPEAKER_01] If you can Google me, [SPEAKER_01] as I always say, [SPEAKER_01] my parents had good SEO. [SPEAKER_01] So, Google Benedict Evans. [SPEAKER_01] And there's a website where [SPEAKER_01] I publish all the presentations that I've done, and you can sign up for my newsletter, [SPEAKER_01] which comes out every week.

1:08:55

SPEAKER_00

[SPEAKER_01] [SPEAKER_00] Otherwise, [SPEAKER_01] how can they be useful to me? [SPEAKER_01] I'm always trying to understand stuff, and I'm always trying to ask different questions. [SPEAKER_01] The worst thing in tech is to [SPEAKER_01] carry on talking about the same stuff. [SPEAKER_01] The moment you really understand something is the moment you have to push on to something else. [SPEAKER_01] And so I'm always trying to think, [SPEAKER_01] no,

1:09:00

SPEAKER_01

am I just talking about the same thing over and over again? Last year, I just spent probably too much time saying, “But these models still hallucinate. Stop telling me they don't hallucinate.” And they do. They still hallucinate. You push them, [SPEAKER_00] [SPEAKER_01] push them a little bit further, [SPEAKER_00] [SPEAKER_01] any question, [SPEAKER_00] [SPEAKER_01] and you'll still get, [SPEAKER_00] [SPEAKER_01] “No, [SPEAKER_00] [SPEAKER_01] that's not true.” [SPEAKER_00] [SPEAKER_01] But that doesn't mean they're not useful. [SPEAKER_00] [SPEAKER_01] So you have to keep pushing, [SPEAKER_00] [SPEAKER_01] keep pushing myself.

1:09:25

SPEAKER_01

[SPEAKER_00] [SPEAKER_01] So that's always the challenge for me: [SPEAKER_00] [SPEAKER_01] How do I push? [SPEAKER_00] [SPEAKER_01] And then, yes, [SPEAKER_00] [SPEAKER_01] if you want me to come and present to your board in the Caribbean, [SPEAKER_00] [SPEAKER_01] then let me know. [SPEAKER_00] And by the way, [SPEAKER_00] the domain is ben- [SPEAKER_00] evans.com. [SPEAKER_00] If folks want to check you out, ben-evans.com. [SPEAKER_00] Thank you so much for being here. [SPEAKER_00] Thanks a lot. [SPEAKER_00] Bye, everyone. [SPEAKER_00] Thank you so much for listening. [SPEAKER_00] If you found this valuable, [SPEAKER_00] you can subscribe to the show on Apple Podcasts,

1:09:44

SPEAKER_01

[SPEAKER_00] Spotify, [SPEAKER_00] or your favorite podcast app. [SPEAKER_00] Also, [SPEAKER_00] please consider giving us a rating or leaving a review, [SPEAKER_00] as that really helps other listeners find the podcast. [SPEAKER_00] You can find all past episodes or learn more about the show at Lenny'sPodcast.com. [SPEAKER_00] See you in the next episode. [SPEAKER_00] See you in the next episode. See you in the next episode. See you in the next episode. [SPEAKER_00] Spotify, [SPEAKER_00] or your favorite podcast app. [SPEAKER_00] Also, [SPEAKER_00] please consider giving us a rating or leaving a review, [SPEAKER_00] as that really helps other listeners find the podcast.

1:10:11

SPEAKER_01

[SPEAKER_00] You can find all past episodes or learn more about the show at Lenny'sPodcast.com. [SPEAKER_00] See you in the next episode. [SPEAKER_00] See you in the next episode. See you in the next episode. See you in the next episode. See you in the next episode.

1:10:22

SPEAKER_00

[SPEAKER_01] See you in the next episode. [SPEAKER_01] See you in the next episode. [SPEAKER_01] Yes, [SPEAKER_01] clearly, there's a bunch of professions where this is a major question. [SPEAKER_01] And particularly if you're an associate or would have been thinking about being an associate, [SPEAKER_01] this is a major question. [SPEAKER_01] And it's very unclear how those professions are going to play out. [SPEAKER_01] It's very unclear what [SPEAKER_01] happens to the pyramid structure of professional services. [SPEAKER_01] The answer, [SPEAKER_01] the only answer I think one can have is: [SPEAKER_01] Don't stick your head in the sand and say,

1:10:38

SPEAKER_00

[SPEAKER_01] "I hate all of this stuff," [SPEAKER_01] because that gives you a great feeling of moral superiority. [SPEAKER_01] And you can go on Bluesky and shout at everybody, [SPEAKER_01] shout at each other about how evil AI is. [SPEAKER_01] Great, [SPEAKER_01] I'm happy for you. [SPEAKER_01] But that's not going to help. [SPEAKER_01] What helps is you diving into this, completely submerging yourself in it, and coming out

1:10:44

SPEAKER_01

understanding what you can do with it, how this changes things, how you can be a great hire. And that may still not help. But if you're going into a law firm and they're saying, "Well, we hired a hundred associates last year, and this year we're only going to hire 50," going to the interview and saying, "Well, I think AI is bullshit, and I'm never going to use it," is probably not the right move. That, that, that may not be particularly comforting, but I don't think there's an alternative. You have to dive into this and absorb it and internalize it and think about what it means, just as you and I did with mobile and with the internet.

1:10:58

SPEAKER_01

[SPEAKER_00] I think that is actually very actionable and [SPEAKER_00] very consistent advice on the podcast: [SPEAKER_00] Just do stuff, [SPEAKER_00] build it, [SPEAKER_00] don't just sit around and pontificate and be [SPEAKER_00] pissed at what's happening. [SPEAKER_00] To [SPEAKER_00] close this out, [SPEAKER_00] I'm going to take us to AI Corner, [SPEAKER_00] a recurring corner of the podcast. [SPEAKER_00] And the question to you is just: What's one way you used AI and use AI in your [SPEAKER_00] work or life [SPEAKER_00] that [SPEAKER_00] is really interesting, [SPEAKER_00] something that other people might [SPEAKER_00] be inspired by? I don't know. I

1:11:25

SPEAKER_01

struggle with this question because I'm the lawyer looking at ChatGPT. So the stuff that I would do that I would automate are precise information-retrieval tasks,

1:11:27

SPEAKER_00

[SPEAKER_01] which is precisely the thing that this is worst at. [SPEAKER_01] And [SPEAKER_01] that's not a criticism. [SPEAKER_01] It's just an observation. [SPEAKER_01] The

1:11:31

SPEAKER_01

stuff that I would want a machine to do for me is the stuff that AI can't do for me very, very, very well at the moment. I use it for proofreading. I use it for images. I used it redecorating my apartment. That worked fantastically well at that. Here's a picture of this room repainted. Add this light and this table and this rug. No,

1:11:42

SPEAKER_00

[SPEAKER_01] change the color of the rug. [SPEAKER_01] There are parts of stuff where it works. [SPEAKER_01] But [SPEAKER_01] a couple of years ago,

1:11:48

SPEAKER_01

somebody said AI is good at stuff that computers are bad at and bad at stuff that computers are good at. And that's— I, I, I struggle to find many, many examples of those where I need it. But then I have a unique, weird job. I sit at my desk all day, trying to synthesize a whole bunch of other stuff into a whole bunch of new ideas. That's not a particularly common way for people to spend their time. I struggle to find AI use cases. I am the accountant looking at the spreadsheet and thinking, "Well, that's very clever. And this is clearly going to completely transform everything. But I actually don't make spreadsheets every day."

1:12:12

SPEAKER_01

[SPEAKER_00] I went to a stand-up comedy show with [SPEAKER_00] Pete Holmes. [SPEAKER_00] I don't know if you know him. [SPEAKER_00] And he made this joke that [SPEAKER_00] we want AI to clean the poop off the street and do all these hard things

1:12:16

SPEAKER_00

that nobody wants to do, but instead, it's, "Oh, let me help you write. Let me help you create imagery." It's this bohemian. It's, "No, I don't want to.

1:12:28

SPEAKER_01

[SPEAKER_00] I don't want to do all these ugly things." [SPEAKER_00] Pete Holmes. [SPEAKER_00] I don't know if you know him. [SPEAKER_00] And he made this joke that [SPEAKER_00] we want AI to clean the poop off the street and do all these hard things [SPEAKER_00] that nobody wants to do, [SPEAKER_00] but instead it's, [SPEAKER_00] "Oh, [SPEAKER_00] let me help you write. [SPEAKER_00] Let me help you create imagery." [SPEAKER_00] It's this bohemian. [SPEAKER_00] It's, [SPEAKER_00] "No, [SPEAKER_00] I don't want to, [SPEAKER_00] I don't want to do all these ugly things. [SPEAKER_00] I want to be creative. [SPEAKER_00] Make art." Yeah. Well, there's, there's, there's

1:12:57

SPEAKER_01

variations of all of this. It's, "I don't want the AI to do the stuff I do for fun. I want it to do the boring stuff that I don't do for fun." Yeah. And finding that mesh— joking apart, this is going to come back to my, my chatbot, chatbot point that the chatbot is a blank screen and a jagged edge. What am I supposed to do, and what will work? And that's a big problem. And the solution to that problem is to wrap it in, in use cases. Part of it is also that AI just disappears. So most of what I write now, I dictate. I dictate it as a voice memo, and that's automatically transcribed. Is that still AI, or is that just voice recognition? Probably an LLM.

1:13:31

SPEAKER_01

There's probably an LLM in there. Okay. So maybe that's AI. Well, okay. So, so what? [SPEAKER_00] At a certain point, [SPEAKER_00] it's just automation. [SPEAKER_00] What do you use for that? [SPEAKER_00] For voice, [SPEAKER_00] voice transcription? So I actually find Apple Notes,

1:13:53

SPEAKER_00

[SPEAKER_01] the app or the one built into the iPhone, works fine. [SPEAKER_01] I'm conscious that people want others,

1:13:57

SPEAKER_01

but I dictate it, though. It is. It worked. So I'm, I'm happy with that. [SPEAKER_00] All right. [SPEAKER_00] Final question before we get to our very exciting lightning round. [SPEAKER_00] Is there anything else that you wanted to share? [SPEAKER_00] Anything else you want to leave listeners with? No, I think I've, I've, I've monologued plenty, and I've gone through a bunch of stuff in the deck. Go read the deck and sign up for my newsletter. And then you will get many more mags of brilliant Benedict Evans wisdom, some of which may even be useful. Somebody answered—

1:14:25

SPEAKER_00

[SPEAKER_01] someone unsubscribed from my newsletter, and they said, [SPEAKER_01] "You didn't, [SPEAKER_01] you didn't give me any actionable stock ideas." [SPEAKER_01] And I'm like, [SPEAKER_01] "Well, [SPEAKER_01] on one level,

1:14:32

SPEAKER_01

that's completely true. On the other level, maybe not." [SPEAKER_00] Well, [SPEAKER_00] with that, Benedict, [SPEAKER_00] we've reached our very exciting lightning round. [SPEAKER_00] I've got five questions for you. [SPEAKER_00] Are you ready? [SPEAKER_00] Sure. [SPEAKER_00] First question. [SPEAKER_00] What are two or three books that you find yourself recommending most to other [SPEAKER_00] people? That's a tough one for me, because I just read an enormous number of books, and then I can't remember which ones I've read. I, I, I sometimes often joke that the classic British comedy from the late 19th century called Three Men in a Boat is my I Ching.

1:14:57

SPEAKER_01

We're having trouble hanging a picture. Well, there's a section about that. We're having trouble doing this. Oh, well, there's a success story about that. All of which are hilarious. So Three Men in a Boat is my I Ching. There's a book by, I think, William Cronon, about the economic history of Chicago, which is fascinating and actually very relevant to technology because it's talking about standardization and packetization and logistics and channel conflict and network dynamics and network neutrality. So when the meatpackers of Chicago reach the point that it's cheaper to ship a cow from New York to Chicago, kill it, pack it,

1:15:21

SPEAKER_01

and then ship it back to New York than to kill it in New York. And the, the pricing of refrigerator cars. And it's exactly like reading about broadband. It's all the same kind of— it's those kinds of business issues,

1:15:30

SPEAKER_00

[SPEAKER_01] which is fascinating.

1:15:30

SPEAKER_01

What else have I read? I don't know. Read books. Read different books. Generally, read books for grown-ups. Please read something other than Lord of the Rings. If you're going to name another company and it's— I saw this, and what was the latest Peter Thiel company? I was like, "Read another book."

1:15:42

SPEAKER_00

[SPEAKER_01] Everything is named after a character from this one book. [SPEAKER_01] There is more than one book in the world. [SPEAKER_01] There is more than one book. [SPEAKER_01] It's those kinds of business issues, [SPEAKER_01] which is fascinating. [SPEAKER_01] What else have I read? [SPEAKER_01] I don't know.

1:15:51

SPEAKER_01

Read books, read different books,

1:15:56

SPEAKER_00

[SPEAKER_01] generally read books for grownups.

1:15:59

SPEAKER_01

Please read something other than Lord of the Rings. If you're going to name another company and it's, I saw this, and what was the latest Peter Thiel company?

1:16:02

SPEAKER_00

[SPEAKER_01] I was, [SPEAKER_01] read another book. [SPEAKER_01] Everything is named after a character from this one book. [SPEAKER_01] There is more than one book in the world. [SPEAKER_01] There is more than one book. [SPEAKER_01] They're all about science fiction. [SPEAKER_01] Read, [SPEAKER_01] read about different things. [SPEAKER_01] Read about things you don't know about. Along those lines, do you have a favorite recent movie or TV show that you've really enjoyed? [SPEAKER_01] I don't know.

1:16:12

SPEAKER_01

I've dropped so badly off the, the, the current media treadmill. And I just spend most of my time watching classics, which are always the ones that you're supposed to have seen and that all seem intimidating. And then you watch them and you're, “Oh, that was actually really good.” I watched The Seventh Seal recently, which is one of those Jake Woody Allen, terrifying, boring movies. And it was brilliant. It was really interesting. And it's, it's only an hour. So go watch, go watch one of those movies that you are supposed to have seen or [SPEAKER_00] hadn't seen. [SPEAKER_00] Favorite recent product that you've recently discovered that [SPEAKER_00] you really love.

1:16:41

SPEAKER_01

[SPEAKER_00] It could be a gadget. [SPEAKER_00] It could be an app. I was speaking at a partner meeting for a company earlier this, this week—what's today? Monday? No, last week. And met the founder of the company, who has a very famous network. The CEO of the company has a very famous name, and I admired his shoes and didn't say anything, but then went and Googled half an hour later. Yeah. Okay. I'll buy a pair of these.

1:16:59

SPEAKER_00

[SPEAKER_01] You want to share the brand, or do you want to keep it,

1:17:01

SPEAKER_01

keep it secret?

1:17:02

SPEAKER_00

[SPEAKER_01] Okay. [SPEAKER_01] We'll keep it secret. [SPEAKER_01] I don't know. [SPEAKER_01] I think it comes in, [SPEAKER_01] it comes in waves of new products, and you [SPEAKER_01] get into waves of new things, and [SPEAKER_01] when's the last time there was a cool app?

1:17:07

SPEAKER_01

iPhone apps, that was— all that white space went. It's partly a function of product shifts, platform shifts. All the white space went for cool new apps. And now we haven't quite got— actually, this is, to the earlier point, we don't have breakout consumer AI apps yet because, I think, because of marginal cost more than anything else, you can't make it free and get 50 million users and then have a revenue model. But we don't have those breakout things yet. For consumers. [SPEAKER_00] Yeah. For consumers. No, I just— we keep getting these ads for voice recorders. Somebody's selling a business-card-size hardware voice recorder. I'm, but I didn't get it.

1:17:49

SPEAKER_01

I've got the voice recorder on my phone. [SPEAKER_00] Yeah. [SPEAKER_00] All kinds of cool stuff coming. [SPEAKER_00] Okay. [SPEAKER_00] Two more questions. [SPEAKER_00] Do you have a favorite life motto that you find yourself coming back to often in [SPEAKER_00] work or in life? I suppose I've mentioned it earlier. Apparently, I mostly say, “It depends.”

1:18:03

SPEAKER_00

That's going to be the title. [SPEAKER_01] It'll probably be okay. [SPEAKER_01] Yeah. [SPEAKER_01] Okay. I don't— that's, that's the vibe I get. I like that.

1:18:13

SPEAKER_01

[SPEAKER_00] I like that. [SPEAKER_00] It's probably going to be okay. [SPEAKER_00] Not for sure. [SPEAKER_00] Okay. [SPEAKER_00] Final question. [SPEAKER_00] I saw somewhere that you own a lot of old phones. [SPEAKER_00] Is that true? It is.

1:18:26

SPEAKER_00

[SPEAKER_01] Yes.

1:18:26

SPEAKER_01

As a— I kept— I was a telecoms analyst and mobile analyst, and I, I kept all my phones up to a point. And now they're uninteresting, but as you may remember, before the iPhone, particularly outside the USA, there was this huge creativity and expansion in what phones looked like, because everyone was innovating around a teeny-tiny gray square. So everyone was trying to differentiate from everything else. Before it results— it's like cars, actually. It's like cars before— before wind tunnels, cars all looked different. And everyone's trying to innovate around because you've got the same four wheels and the same engine.

1:18:58

SPEAKER_01

And everyone's trying to differentiate based on the shape, and then everything converges on one shape. And it's the same with phones. Everyone— everything converged on one shape. Before that, it was all this innovation. So yeah, I, I have a whole bunch of PDAs and, It's like cars before street, before wind tunnels. Cars all look different. And everyone's trying to innovate around because you've got the same four wheels and the same engine. And everyone's trying to differentiate based on the shape,

1:19:15

SPEAKER_00

[SPEAKER_01] and then everything converges on one shape. [SPEAKER_01] And it's the same with phones. [SPEAKER_01] Everyone— [SPEAKER_01] everything converged on one shape. Before that, [SPEAKER_01] it was all this innovation. [SPEAKER_01] So, yeah, [SPEAKER_01] I, [SPEAKER_01] I have [SPEAKER_01] a whole bunch of PDAs and [SPEAKER_01] smartphones. And how many phones are we talking about? [SPEAKER_01] I don't know, 20 or 30. Okay. Okay. It's not so crazy. What's the oldest one? What's the oldest one you got? [SPEAKER_01] So, I have one of those— [SPEAKER_01] I should have easily told me I'd have got the box down. [SPEAKER_01] I have one of those Ericsson

1:19:48

[SPEAKER_01] shark-fin flip phones from '98 [SPEAKER_01] or something, [SPEAKER_01] which is very, [SPEAKER_01] again, [SPEAKER_01] hardware design, [SPEAKER_01] visual design, [SPEAKER_01] trying to differentiate. [SPEAKER_01] I've got an i-mode phone from 2001 and a J-Phone phone from 2001 that has a camera. [SPEAKER_01] So, I came back from Japan in 2001, and I found it had a color screen and a camera. [SPEAKER_01] And [SPEAKER_01] I just had endless client meetings, and people just wanted to see the phone with a color screen. [SPEAKER_01] It was a mind blow. [SPEAKER_01] It didn't work outside Japan. [SPEAKER_01] Actually, [SPEAKER_01] I plugged it in the other day.

1:19:50

[SPEAKER_01] It still charges up. [SPEAKER_01] Clearly, I can't do anything with it. [SPEAKER_01] And [SPEAKER_01] there's— [SPEAKER_01] there's a little bit of an analogy in there as well. [SPEAKER_01] And [SPEAKER_01] we thought there'd be all these different shapes and sizes. [SPEAKER_01] And before the iPhone, [SPEAKER_01] people imagined, [SPEAKER_01] well, [SPEAKER_01] some people will have a little pocket PC, and some people have a keyboard, and you have folding, [SPEAKER_01] or there were all these different ideas for what it would look like. [SPEAKER_01] And it all— [SPEAKER_01] we didn't realize it was all going to converge on one device. [SPEAKER_00] Benedict.

1:19:50

[SPEAKER_00] This was amazing. [SPEAKER_00] I learned a ton. [SPEAKER_00] I feel better after this conversation. [SPEAKER_00] Two final questions. [SPEAKER_00] Where can folks find you online? [SPEAKER_00] Where do they find this presentation? [SPEAKER_00] And how can listeners be useful to you? [SPEAKER_01] You can Google me. [SPEAKER_01] As I always say, [SPEAKER_01] my parents had good SEO. [SPEAKER_01] So, Google Benedict Evans. [SPEAKER_01] And so there's a website where [SPEAKER_01] I publish all the presentations that I've done, and sign up for my newsletter, [SPEAKER_01] which comes out every week. [SPEAKER_00] Otherwise, [SPEAKER_01] how can they be useful to me?

1:19:50

[SPEAKER_01] I'm always trying to understand stuff, and I'm always trying to ask different questions. [SPEAKER_01] The worst thing in tech is to [SPEAKER_01] carry on talking about the same stuff. [SPEAKER_01] The moment you really understand something is the moment you have to push on to something else. [SPEAKER_01] And so I'm always trying to think, [SPEAKER_01] no, [SPEAKER_01] am I just talking about the same thing over and over again? [SPEAKER_01] Last year, [SPEAKER_01] I just spent probably too much time saying, [SPEAKER_01] but these models still hallucinate. [SPEAKER_01] Stop telling me they don't hallucinate. [SPEAKER_01] And they do.

1:19:50

[SPEAKER_01] They still hallucinate. [SPEAKER_01] You push them— [SPEAKER_01] push them a little bit further, [SPEAKER_01] any question, [SPEAKER_01] and you'll still get, [SPEAKER_01] no, [SPEAKER_01] that's not true. [SPEAKER_01] But that doesn't mean they're not useful. [SPEAKER_01] So you have to keep pushing, [SPEAKER_01] keep pushing myself. [SPEAKER_01] So that's always the challenge for me: [SPEAKER_01] how do I push? [SPEAKER_01] And then, yes, [SPEAKER_01] if you want me to come and present to your board in the Caribbean, [SPEAKER_01] then let me know. [SPEAKER_00] And by the way, [SPEAKER_00] the domain is ben-evans.com.

1:19:50

[SPEAKER_00] If folks want to check you out, ben-evans.com. [SPEAKER_00] Thank you so much for being here. [SPEAKER_00] Thanks a lot. [SPEAKER_00] Bye, everyone. [SPEAKER_00] Thank you so much for listening. [SPEAKER_00] If you found this valuable, [SPEAKER_00] you can subscribe to the show on Apple Podcasts, [SPEAKER_00] Spotify, [SPEAKER_00] or your favorite podcast app. [SPEAKER_00] Also, [SPEAKER_00] please consider giving us a rating or leaving a review, [SPEAKER_00] as that really helps other listeners find the podcast. [SPEAKER_00] You can find all past episodes or learn more about the show at Lenny'sPodcast.com. [SPEAKER_00] See you in the next episode.

1:19:50

[SPEAKER_00] See you in the next episode. See you in the next episode. See you in the next episode. See you in the next episode.

Reading tools

Type to find a passage

Appearance
Ask this transcript

Notebook

Export Markdown

Add a note