I feel like I’ve been hearing this stupid “PhD level intelligence” claim about every LLM that’s come out since ChatGPT was first released, including GPT-3.0 which it launched with. It kind of amazes me that people keep falling for it and not questioning how the new model having “PhD level intelligence” is both a true claim and also noteworthy when the claim is made about every new model.
It will analyze and parse primary sources with all the discernment of a pure math PhD! Design bridges with all the insight of a literature PhD! Diagnose medical problems with all the experience of a supreme court justice!
Iv met enough phd holders to know that they can and frequently are still unabashedly wrong on the vast majority of everything they talk about that isn’t hyper specific to a narrow and niche topic.
So phd level intelligence to me just means it’s more prone to the being confidently wrong and judgemental.
I feel like I’ve been hearing this stupid “PhD level intelligence” claim about every LLM that’s come out since ChatGPT was first released, including GPT-3.0 which it launched with. It kind of amazes me that people keep falling for it and not questioning how the new model having “PhD level intelligence” is both a true claim and also noteworthy when the claim is made about every new model.
It will analyze and parse primary sources with all the discernment of a pure math PhD! Design bridges with all the insight of a literature PhD! Diagnose medical problems with all the experience of a supreme court justice!
Iv met enough phd holders to know that they can and frequently are still unabashedly wrong on the vast majority of everything they talk about that isn’t hyper specific to a narrow and niche topic.
So phd level intelligence to me just means it’s more prone to the being confidently wrong and judgemental.
sometimes they are outdated if they have been doing projects or research for a while too.