Jul 1, 2024 · 37m · another-podcast
The AI summer
0:00 / 0:00
gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions
Tech analyst Benedict Evans and co-host Toni Cowen-Brown examine the state of generative artificial intelligence, analyzing Apple Intelligence, empirical enterprise adoption metrics, interface integration models, and the economic sustainability of massive infrastructure investments.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. The hosts hold 80.1% of the talking time here. How this is scored →
The hosts as informed peer 8.7 Guest teaching 1.1 Guest disagreement 1.0 The hosts pushing back 1.4
speaking balance: gold is the hosts, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 11:18 Challenging the teasing over mobile ChatGPT adoption
Cowen-Brown pushes back assertively when Evans jokingly suggests editing out her admission that she only recently downloaded the mobile ChatGPT app.
Hardest push from the hosts ▶ 33:53 Pushing back on Nvidia as guaranteed winnerEvans immediately nuances Cowen-Brown's claim that Nvidia is the undisputed winner, stressing that this is only true for now and detailing vulnerability to hyperscaler CapEx deceleration.
Biggest teaching moment ▶ 21:00 Explaining public confusion around the AI-designed trophyCowen-Brown provides on-the-ground insight from the Canadian GP, educating Evans on how regular consumers still fail to distinguish between physical fabrication and AI digital design.
The host holds their own ▶ 6:40 Historical analysis of 1987 InfoWorld spellchecker marketEvans displays encyclopedic tech history knowledge, citing 1987 standalone software matrices to prove that generative text features will inevitably become standard OS infrastructure.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | The hosts as informed peer | Guest teaching | Guest disagreement | The hosts pushing back | Why |
|---|---|---|---|---|---|---|
| Catching Up and Setting the AI Agenda | 8 | 1 | 1 | 1 | Evans immediately takes charge of structuring the episode, articulating a comprehensive taxonomy of AI questions separated into insider tech dilemmas and external enterprise adoption. Cowen-Brown collaborates naturally, tossing in occasional framing questions about the end user. | |
| Features vs. Native Products: Spellcheckers to Firefly | 9 | 1 | 1 | 1 | Evans draws deeply on computing history, citing his archival research of 1987 InfoWorld spellchecker software comparisons to explain how stand-alone software collapses into OS features. Cowen-Brown supplements with modern examples like CapCut inside TikTok. | |
| Mobile AI Utility and Monaco Travel Anecdotes | 9 | 1 | 1 | 2 | After brief banter regarding travel and mobile app adoption, Evans breaks down detailed enterprise survey datasets from Reuters, Deloitte, Bain, and Accenture's quarterly earnings metrics. Cowen-Brown listens and clarifies terms like production deployments. | |
| Public Perception, Hallucinations, and Sports AI | 8 | 3 | 2 | 2 | Cowen-Brown introduces the Canadian GP generative AI trophy controversy to illustrate broad public misunderstanding of AI. Evans contextualizes this with hallucination metrics from Deloitte surveys and the shifting discourse at Davos. | |
| The Four Levels of Generative AI Integration | 9 | 0 | 0 | 0 | Cowen-Brown invites Evans to detail his four-level framework of generative AI integration. Evans lays out the model cleanly from background OS features up to autonomous oracles, maintaining total pedagogical authority. | |
| Disruption Frameworks and the Oracle at Delphi | 9 | 1 | 1 | 2 | Evans critiques innovation consulting frameworks using the Oracle at Delphi metaphor and the historical misjudgment of Airbnb and the iPhone. Cowen-Brown engages constructively with how infrastructure shifts redefine business questions. | |
| Nvidia Capital Flows and Moving Target Realities | 9 | 1 | 1 | 2 | Evans analyzes Nvidia's $25B quarterly sales, questioning defensibility and explaining why current AI represents a moving target unlike smartphones in 2008. The hosts close the discussion in complete alignment. |
Statements from this episode (14)
Evans: High inference costs prevent free 100-million-user consumer AI apps
“Because at the moment you can't make a consumer app that's free and have a hundred million users because you can't afford the inference cost.”
Evans: Tech incumbents always attempt to turn new technology into features
“The incumbents always try to make the new thing a feature.”
Evans: Rabbit's 'cack-handed' AI hardware is becoming a 'smoking hole'
“And we saw this kind of completely cack-handed attempt at doing this from Rabbit, which is now well on the way to being a smoking hole in the ground.”
Evans: 33% to 50% tried ChatGPT, but daily active users are 1% to 5%
“And the top line is basically that in a bunch of developed world countries, something between a third to a half of people have already tried ChatGPT.
And so that's a very glass half full kind of statement.
On the other hand, the number of people you've turned …”
Evans: Accenture reached nearly $4B run rate in generative AI work
“The May quarter over 900.
So they're now at close to a four billion dollar run rate in doing, except doing generative AI work for their clients.”
Evans: $1M enterprise consulting spend buys GenAI development, not deployment
“Because a million dollars gets you developing, it does not get you deploying, even from, even at Accenture's rates.”
Evans: Davos attendees stopped discussing AGI between 2023 and 2024
“Last year, everyone at Davos was talking about AGI. This year, no one at Davos was talking about AGI.”
Evans: Deloitte survey shows one-third believe generative AI is always accurate
“They asked about hallucinations, and they said, do you believe that generative AI always produces accurate answers? And the question is still, like, a third of people say yes, even if they've tried it.”
Evans: 99% of crypto and NFTs were useless scams
“It wasn't all a scam. Just 99% of it was a scam and useless.”
Evans: Probabilistic AI is a feature, not a standalone product
“It seems to me like the Occam's razor is, if you can't guarantee it's right, then it's features. It has to be wrapped in something that can manage or control or shape expectations or something around what it is that this is doing, because you can't just use it…”
Evans: Technologies transition over time from 'AI' to 'smart' to ordinary software
“Things go from AI to smart
To just software. There's a period in between where it's smart, and then it's just software.”
Evans: Every large enterprise is exploring AI, but none are in production
“It seems like we're now sort of at the point that every big company's got this, everyone big company sort of is thinking about what they do with it, and no one's got any production yet.”
Evans: OpenAI will face half a dozen competitors
“I'm not sure, you know, there would be a big debate about whether OpenAI is worth anything right now, or what it's worth based on the fact that there's clearly going to be half a dozen competitors to it.”
Evans: Enterprises require dedicated product UI over raw ChatGPT prompts
“Ask yourself why it is it's a lot easier to go to a company and say, here's a slide maker, than to go to a company and say, well, look, ChatGPT is just multimodal, so just ask what you want. Tell it the slides you want. No, you want UI. You want product around…”