I don’t understand what you’re saying. You know the AI is incapable of reading your mind, right? Can you provide more information?
You'd be surprised at just how much data you can pry out of an LLM that was merely exposed to a single long conversation with a given user.
Chatbot LLMs aren't trained to expose all of those latent insights, but they can still do some of it occasionally. This can look like mind reading, at times. In practice, the LLM is just good at dredging the text for all the subtext and the unsaid implications. Some users are fairly predictable and easy to impress.
This extends even to novel and unseen tests - so it's not like they could have memorized all of them.
Base models perform worse, and with a more jagged capability profile. Some tests are easier to get a base model to perform well on - it's likely that they map better onto what a base model already does internally for the purposes of text prediction. Some are a poor fit, and base models fail much more often.
Of course, there are researchers arguing that it's not "real theory of mind", and the surprisingly good performance must have come from some kind of statistical pattern matching capabilities that totally aren't the same type of thing as what the "real theory of mind" does, and that designing one more test where LLMs underperform humans by 12% instead of the 3% on a more common test will totally prove that.
But that, to me, reads like cope.
Of course they can, just like a psychiatrist can.
There was nothing involved like what we refer to as "AI" today.
Use the LLM more until you are convinced. If you are not convinced, use it more. Use it more in absurd ways until you are convinced.
Repeat the above until you are convinced.
So he's not just an LLM evangelist, he also writes like one.