There are many different types of AI. One way to break it down further is into the Generative AI (GenAI) and Predictive AI categories. Generative AI gives the most impressive demos, writing poems, songs, making interesting pictures etc. GenAI is what got most people paying attention to AI. GenAI is turning out to have limited economic use.
Predictive AI, on the other hand, is proving to be very useful and the use cases are broadening and deepening. For example, the two examples in this article (travel planning, automotive finance) are both predictive AI. AI that dramatically speeds and augments decisions that skilled humans would make.
I see it as a valuation problem and a moat problem. Nvidia at $4.4tn at 16% of US GDP. Yet Nvidia is about to start facing massive competition and will lose share and pricing power. Open AI is getting crucified by Google, and could go bust. Small models are just as good as large models, and they can be run at home outside the cloud. Open source is just as good as closed source. The technology is amazing, but is totally democratised, and is free. So very few companies that provide the technology will be able to retain any value out of it. No moat. Good for the economy, bad for the S&P. Which says Long SPW/ Short SPX
I was not clear. What I meant to say is LLM's have no moat, as far as I can tell. And so this ought to be a major problem for their profitability over time. And this fact will eventually end the (temporary) upstream moat that Nvidia currently enjoys.
If AI can tell us what the profit margin of a given good or service needs to be and does so accurately, isn't that inflationary in that no one would take on new work unless it was profitable for them? In recessions many businesses take on work for less than cost, to keep the lights on. The perverse reality of using AI to cost things will have unusual side effects leading to inflation in the future.
There are many different types of AI. One way to break it down further is into the Generative AI (GenAI) and Predictive AI categories. Generative AI gives the most impressive demos, writing poems, songs, making interesting pictures etc. GenAI is what got most people paying attention to AI. GenAI is turning out to have limited economic use.
Predictive AI, on the other hand, is proving to be very useful and the use cases are broadening and deepening. For example, the two examples in this article (travel planning, automotive finance) are both predictive AI. AI that dramatically speeds and augments decisions that skilled humans would make.
good point
I see it as a valuation problem and a moat problem. Nvidia at $4.4tn at 16% of US GDP. Yet Nvidia is about to start facing massive competition and will lose share and pricing power. Open AI is getting crucified by Google, and could go bust. Small models are just as good as large models, and they can be run at home outside the cloud. Open source is just as good as closed source. The technology is amazing, but is totally democratised, and is free. So very few companies that provide the technology will be able to retain any value out of it. No moat. Good for the economy, bad for the S&P. Which says Long SPW/ Short SPX
All possible - but re the moat - when capacity is constrained, everyone has a moat!
I was not clear. What I meant to say is LLM's have no moat, as far as I can tell. And so this ought to be a major problem for their profitability over time. And this fact will eventually end the (temporary) upstream moat that Nvidia currently enjoys.
Fair point - with so many well capitalised competitors - maybe its a long term price war? Somehow tech companies seem good at avoiding that outcome.
If AI can tell us what the profit margin of a given good or service needs to be and does so accurately, isn't that inflationary in that no one would take on new work unless it was profitable for them? In recessions many businesses take on work for less than cost, to keep the lights on. The perverse reality of using AI to cost things will have unusual side effects leading to inflation in the future.
The robot in the AI generated picture in your article looks like a sinister version of Ultraman
Good spot - AI generated of course. I am a big fan of ultraman
Insightful article Russell !