How would AI do that?

I’m genuinely a little bit more productive with ChatGPT. A lot of tasks I do take 30 mins or a hours less, several times per week.

So a basically trivial improvement in other words.

Half an hour of time recovered using a system not specialized to the work at hand. With better tooling and integrations that could scale to several hours a day spent doing more valuable work.

How is that trivial? Even one task done a few times a week taking half an hour less is at least a 2% increase in productivity (assuming these tasks are work related and a 40 hour work week). "A lot" of such tasks would easily add up to 10%+.

I believe his statement lies in the fact that wealth is not money, but productivity.

He's saying if every atom of the economy is n% more productive, that will grow the economy by n%.

With the first generation of tools.

If there's a sure-fire early way to make money from AI, it's going to involve betting against people who blow it off.

I don't blow off the future. But I do think there's a lot of excess hype currently.

I save at least 10 hours a week using ChatGPT. Likely more.

Automating knowledge work -> increasing labor productivity.

Absolutely no chance that AI is able to do that. Investing decisions are too subjective for the current state of AI. A good value investor knows this… and also knows that you don’t need a bull market to make decent returns.

> Investing decisions are too subjective for the current state of AI

I think they are talking about AI driving growth in the broader economy, not making investment decisions.

Even still, it's not clear that AI will drive that kind of growth.

It is not clear exactly how, but kind of makes common sense.

Productivity is about producing the same things faster. If AI can do part of our work for us it means we can accomplish more with the time saved.

Or, fewer and fewer people are needed for approximately the same amount of production. This may lead to a further bifurcated society.

I do think that a service-oriented economy has many, many more nooks and crannies to hide non-productive jobs, though, and that could perpetuate the “bullshit job economy” hypothesis.

Fewer people are needed to perform a specific job as we do it today. But there will always be room to improve the products and services we produce and to make them cheaper.

Consider the steam-locomotive. It was a great invention but we didn't stop there now we have bullet-trains and hyperloops.

There is also plenty of room at the bottom as famously noted by Feynman I think. We can make products smaller to make them easier to carry around. New cheaper price points create new demand and new markets.

Also as economies improve, people will have fewer children, because they don't need them to take care of themselves at old age. So it's not like we will have too many people having nothing to do.

On the contrary I think AI can be the great equalizer: Rather than having fewer people who know how to do things, we will have more people who are able to do things with the help of AI.

The technology by itself will not produce a great society however. We need democracy to accomplish that. As people will have more free time they will have time to study and more and more people will demand democracy because they will understand it's the only way to avoid wars.

Yes, exactly. And agreed, it’s not clear. (Hopefully I conveyed wide error bars, my whole point is that certainty seems unjustified at this point in time.)

The number of people on planet earth who deeply understand the current state of the art in LLMs, the available information in a format an LLM can "reason about", good ways to condense information so that X v Y decisions can be made across a lot of Xs and Ys, and deeply understand how to do value investing in a practical way and enough about a set of industries to pull it off, is about zero.

Regardless, I'd happily take a bet at even money up to a reasonable sum that we'll wake up in 20 years to find that a system using LLMs has just destroyed the market doing fundamental analysis, for 15 years.

It's not even that, the way markets work is inherently unpredictable, in a self fixing way.

If someone finds a way to actually predict the market, then in taking advantage of that they will add unpredictability into it again. This is part of the entire basis of how markets work!