AN EXPERT VIEW ON AIIn this edition, Tech in Asia caught up with well-known analyst Benedict Evans on the sidelines of the SuperAI event in Singapore, where he presented an updated version of his "AI eats the world" presentation. He told us that the industry's biggest shift over the past six months has been a growing divergence in product strategies among companies and a clearer understanding of how people will use AI beyond the model labs. This interview has been edited for brevity and clarity.  How did you come up with the "AI eats the world" narrative? It's a reference to the essay that Mark Andreessen wrote called Why Software is Eating the World. Mobile happened, now what? Now it's clear that it's AI and this is sort of the center of the tech industry for the next 10 to 15 years, maybe much more than that. AI is sort of the new central gravity of tech. However, one should be slightly conscious of all the stuff that we were excited about before AI is still there. Between your two "AI eats the world" presentations, what changes do you think were the most significant in the industry? From late 2022 to the middle of last year, there are four basic questions that haven't really changed: Will the models get bigger, how much bigger, how expensive does this get, how to scale? But you couldn't really see the differences in product strategy - what are the kind of business questions as opposed to the science questions. And now I think you have much more clarity on a divergence between different companies on questions of how people use AI, what this means for the software industry outside the model labs, how companies might be thinking what this will do for them, how it will change things. What are people getting wrong when they claim AI will lead to mass unemployment? Every new technology creates new jobs and uncreates jobs. That was true of the internet, obviously, and it was true of PCs, and it was true of computers. This is a lump of labor fallacy, which any first-year economics student will tell you about. Each of those waves of technology creates pain: people lose their jobs, job categories go away, industries disappear, companies disappear, and that is painful. That's different from mass unemployment because in the end, net employment didn't change. You wrote a long essay about OpenAI and the challenges it may face over the long term. What do you think the media is getting wrong about OpenAI? I think the challenge is that it's hard to see how having a foundational model by itself has any sort of fundamental competitive advantage. There's no mechanism that would mean you have something no one else could do. If you think about everything from Google Search to Instagram to the iPhone, you have these very strong winner-takes-all effects. You have a user base and network effects that mean even if other people do it, they can't catch up. We don't see those mechanics in foundational models today. The model itself doesn't appear to have really strong competitive differentiation. The question is: What do you build around it? It looked like last year OpenAI's answer was, "We're just going to try and build a lot of things on top and see which ones work." Anthropic's answer was, "No, we're just going to do coding." As an analyst, do you feel a need to remain neutral in your assessment of AI when there is so much hype around it at the moment? The answer to most things is never clear. There's never one correct answer. Sometimes, there are narratives that I think are based on a misunderstanding of the technology or a misunderstanding of human behavior. Sometimes, there are arguments that are based on different attitudes toward how society should work. The most important thing in politics or life is to accept that other people can have different opinions from you, and that's OK. |
