Greg Jensen, Bridgewater’s chief information officer, spoke with Bloomberg, saying that large language models like OpenAI’s chatbot are “awful” at reading the market.
“If someone is going to use large language models for stock picking, I think that’s hopeless — that’s a hopeless path,” he said.
Bridgewater does have a benefit though:
- “Combining large, poorly accurate language models with statistical models that are good for the past but poorly accurate for the future.” “Bringing those together, you start to build an ecosystem that I think can achieve the kinds of things that Bridgewater can do – but it can do it on a much larger scale.”
He added that if you have “the ability to control hallucinations and errors” you can do “a tremendous amount at a rapid rate”.
He’s right about “hallucinations and mistakes” on this stuff, yeah. If you find yourself on the other end of the trade with one of them, it’s probably time to double down 😉