It used to be that when I used AI I found that it would give me obviously wrong answers to questions about things I had some knowledge about. Later new versions would arrive and I would give it another chance. I found that it was getting better. While still giving incorrect information, it would offer some things that were useful that I might not have considered. But nuance and context were simply not there. So I thought it was useful for bouncing some ideas off of, as long as I was very careful to double-check what it was telling me. I understood that if I went pubic with any conclusions based on AI I was risking my reputation.

AI Got Better

Lately it has gotten much better. It can go into some depth, with some nuance and a bit of understanding of context. It seems to be giving fairly comprehensive answers that look right. I could draw some conclusions from this if I accepted what I was seeing.

BUT then I notice the information in the response is not quite right. So I point this out to the AI, offer some more sources to check, and it updates its answer. The update provides responses that look even more informed, with additional nuance and context, but lead toward different conclusions from its previous response. If I dive into those responses they still aren’t quite right, leaving out inputs that would alter the response.

I find I can continue this cycle for some time. Each time I get informed responses that look good, and from which I can draw conclusions. If I’m not careful and skeptical I am risking drawing the wrong conclusions. If I am public with those conclusions I am risking my reputation.

This Is Just Confirmation Bias

I have realized that what I am experiencing, even with the greatly refined AIs that are out there now, is a confirmation bias machine. I can continue to refine my prompts and the AI will continue to refine its responses, until I am satisfied with what I am seeing and draw conclusions. But I could also continue this cycle. Until I am satisfied with what it is telling me.

This isn’t about being right, it’s about seeing responses that satisfy people that they are correct. But all AI is doing, still, is telling people what they want to hear. It’s just doing a better job of it! AI is not “getting better.” It is getting better at making you think it’s getting better. This is about accepting conclusions, not about drawing correct conclusions.

And THIS is what investors and businesses and governments and others are already using to gather information and come to conclusions and act.

Risking the World

The Fox News hosts running the US government are using drawing conclusions based on what AI is telling them. I would be risking my reputation if I accepted that. They are risking the world.