Discussion about this post

User's avatar
Gerben Wierda's avatar

Oh, this is too good. "[W]e hypothesize that ToM-like ability emerged spontaneously and autonomously, as a byproduct of models’ increasing language ability.". The phrase 'seeing Jezus in toast — the science version' springs to mind.

Expand full comment
Spherical Phil - Phil Lawson's avatar

If GPT were to get (or compile) an accurate explanation of theory of the mind, while impressive, that would not mean that GPT will be able to safely counsel a person. I have been creating online tech applications for psychologists, therapists, counselors, coaches etc. for 17 years. Tech designed to be used engaging a real human person about their life and then actually aiding them to deal with issues and situations they face requires advanced reasoning. Especially since the real person in many cases is not capable of accurately and calmly explaining their situation or problem in clear, simple to understand language. And more importantly any intelligent tech would require a way to at the very least emulate some form of compassion or care for the person, which requires ‘knowing’ the person, not theory of mind. These are not abilities, skills, or qualities possible in the current design and structure of LLMs or in ML.

Expand full comment
34 more comments...

No posts