James Zou: one of the most prolific and creative A.I. researchers in both life science and medicine

A podcast with James Zou, a Stanford computer scientist who is lighting it up in life science and medical A.I. Recent papers discussed in this podcast include degradation of GPT-4 performance, using large language models for peer review, and taking more than 200,000 pathology posts from Twitter to build a visual-language model.
For more from Ground Truths, visit: erictopol.substack.com/

[@CMSonYT]

the reason its becoming worse is the same reason it's getting "safer" It's anything but obscure that when people from GPT3 to GPT4 that it became "safer" but also less useable for a lot of people

[@NegativeBodhiImage]

The safer it gets, the worse it gets, the dumber it gets, the less reliable it gets and the less logical it becomes. That should tell you everything you need to know right there, but instead Zou says "getting safer, which is good/better". The very notion that hearing something you dont like, is "unsafe" to begin with is at the very core of this problem. What is more unsafe is believing things that are fallible based on erroneous assumption.