AI self-reinforcement

Forgive me for the next little while as I am seemingly writing more about epistemology than finance, but trust me, it is all related!

The algorithms for some reason decided to put on my feed this video of Senator Bernie Sanders talking to Claude.

I found it fascinating, for reasons not intended by the producers of the video.

The intended payload of this video is Sanders trying to convince his audience that curbs are required on AI data centre development with the added irony of Claude AI agreeing with him at the end.

I don’t know if this is a good or bad idea. Indeed, for the purposes of this post, it doesn’t matter.

However, what I was focused on was the susceptibility of confirmation bias on the user of AI.

At 5:34 in the video clip, Sanders asks:

“… would you support, or think it’s a good idea, to have a moratorium on the development of new AI data centers?”

Claude answers (paraphrasing), there are trade-offs. Perhaps a half-way solution will work.

Sanders, at 7:05, adds a couple pieces of evidence in support of his position.

Claude: “You’re absolutely right, Senator!”

Where am I going with this?

More and more people are using AI tools to analyze external circumstances and make portfolio selections. AI will happily take the information and validate the input of speculations of the user. An unperceptive user will take the rendition as gospel and apply it accordingly.

It is very easy to feed quarterly financials, MD&A statements and conference call transcripts and ask AI to perform reconciliations or summarize highlights and the like. The subtle touches missing from this process is very similar to language translation – while it can deliver the general message, there will be nuances and context missing that are not included in the petabytes the AI machine combs through when delicately lacing the words it chooses to interface your brain with.

This is where competitive advantage can be exploited. Absent of the correct nuance and context lies opportunity. In particular, if too many people (and their AI agents self-reinforcing such thought) believe something in one direction, markets will stretch in one direction until circumstances change perception – sometimes taking a lot longer than most will ever anticipate.