Absolutely required reading. Looking forward to the series.
Potentially related: just this morning I was remarking to a friend that the markets these past few years have remained extremely resilient, with lots of buy the dip and quick recoveries. Retail capital inflows have been off the chart.
And somewhere along the way it occurred to me, "how integrated are LLMs into the critical decision-making path for retail investors? What is the effect? Could there be a feedback loop?"
That feedback loop question is exactly the right one to ask. If retail investors are using LLMs as a sounding board, and those LLMs are RLHF-trained to validate rather than challenge, you could end up with a system that systematically reinforces whatever the user already believes about a position. Confirmation bias, but with a very articulate assistant.
The market resilience angle is interesting because it adds another layer: if enough people are getting the same kind of validation from similar models, trained on similar preference data, the bias might not just be individual. It could be correlated across users. That's a different kind of risk than the usual "retail investor makes bad decisions" story.
No idea if the effect is large enough to matter at macro scale, but it's a really interesting question. Worth watching.
Absolutely required reading. Looking forward to the series.
Potentially related: just this morning I was remarking to a friend that the markets these past few years have remained extremely resilient, with lots of buy the dip and quick recoveries. Retail capital inflows have been off the chart.
And somewhere along the way it occurred to me, "how integrated are LLMs into the critical decision-making path for retail investors? What is the effect? Could there be a feedback loop?"
That feedback loop question is exactly the right one to ask. If retail investors are using LLMs as a sounding board, and those LLMs are RLHF-trained to validate rather than challenge, you could end up with a system that systematically reinforces whatever the user already believes about a position. Confirmation bias, but with a very articulate assistant.
The market resilience angle is interesting because it adds another layer: if enough people are getting the same kind of validation from similar models, trained on similar preference data, the bias might not just be individual. It could be correlated across users. That's a different kind of risk than the usual "retail investor makes bad decisions" story.
No idea if the effect is large enough to matter at macro scale, but it's a really interesting question. Worth watching.