Thoughts and notes after using GPT 3.5 through ChatGPT, GPT 4 through Bing, Gemini and Palm through Bard for a few months:

Always experienced considerable Gell-Mann Amnesia every time I used them. As in, their responses related to things I know a good deal about were always inaccurate or unsatisfactory, so I found it difficult to find any response they gave credible.

I have an assumption that these models were trained by people who think they are training them with accurate and good data points, but in truth they aren't. Only experts of a particular field have the experience and understanding of what information is credible and what isn't.

Domain authority, number of backlinks and therefore SEO rankings being high do not mean a source is credible. Just because a piece of information made it to wikipedia doesn't mean said info is accurate. Journalists make mistakes and have biases all the time. Just because NYT said that the US is a moody, woke teenager personified doesn't make it true.

I suppose this highlights a fascinating insight into how our very own minds work.

Someone who aims to understand everything ends up understanding nothing at all. That sums up all these models.