@garymarcus A big problem I have seen is that they can justify anything you want them to justify, very confidently!

I have still found them to be useful to do mundane tasks like cleaning up emails.

@andreacpa0 @nntaleb

Helium https://www.helium.com/ is a great example of what it could be.
As 5G starts to roll in, and all phones will have a lot of shareable bandwidth, I see it quite useful.

@paulg Similar to writing proofs... Get stuck when the underlying idea has issues with it.

This is a place LLMs have to improve. Right now, LLMs would make anything fit!

@tao A potential way to counter this, which I am planning to use in my class, is to create questions that lead to wrong answers from chatGPT and evaluate students based on whether they are able to catch the mistake.

It will be a version of the multiple choice questions.