Alex Good (@goodalexander) on twitter, talks about this concept of a “hallucination yield”. It’s the idea that LLMs favor certain stocks or assets unreasonably based on their training data.

@goodalexander: “im a recently converted fan of “hallucination yield” - the idea that ChatGPT arbitrarily likes certain stocks or thinks they’re bigger than they are. hard to backtest bc the training cut offs but … yea. works.”

He’s right though, it works. (And by the way this idea may have come from someone else but I couldn’t find the original source.)

Ask ChatGPT about what sectors should do well under the new incoming Trump administration. For some reason it calls out Small Modular Nuclear Reactors, and then it will go on to recommend the stock “NuScale Power” ($SMR).

Want to see that chart?

Now, NuScale Power is a real company. It’s probably the most obvious one to recommend. But it’s not the only one. It’s not even a particularly good one based on various metrics (valuation, financials, etc).

How about cryptocurrencies? Or specifically “memecoins” (that topic is breaking out in google trends right now).

It’s going to invariably talk to you about Dogecoin.

LLM Market Premium

I don’t know that I would call this “hallucination yield”. It’s not that it’s hallucinating, its just that it’s been trained on a lot of data and it’s going to favor certain things.

I think it’s more likely that this is a form of market premium.

LLM Market Premium - to me that is the correct term for this phenomenon. To take this into account you need to think about when this is going to have a lot of weight. What situations are going to be likely to see outsized market impact induced by LLMs?

Greatest impact / how to use this.

1. You need to focus only on ChatGPT. Other LLMs are just not that popular.
2. Use the free / or most basic tier of ChatGPT.
3. Focus on events that retail investors are likely to buy into.
4. Don’t dig very deep, just get what the most basic preferences of ChatGPT are.

As we know, this kind of market participant behavior ends up being reflexive. If money is concentrated in one area, it’s going to get more attention. And that’s going to drive more money into it. There is nothing so predictable as human greed.

I have a lot more to say about the younger generations hypergambling behavior. But that’s a topic for another day.