Raw LLM Responses

> Gen AI is a fancy auto-complete, and is fundamentally flawed. There's no science that can be done with Gen AI. Mathematician here. This is not accurate. In math, high quality LLMs have been used to make mathematical discoveries both in developing new proofs and in finding counterexamples. Mathoverflow has been compiling [a list of major examples of this](https://mathoverflow.net/questions/502120/examples-for-the-use-of-ai-and-especially-llms-in-notable-mathematical-developme) which is now so long that there's discussion of just stopping making the list longer. AI generated results like this need to be checked very carefully, but it does work. This isn't the only way scientists are benefiting from LLMs. For example, they can be used to help find obscure terms in literature. If one doesn't know what term is used for some definition or if it is in the literature, you can ask the AI. One gets a term from it and a few references. One does of course need to then check that the references are not hallucinated, and that the term means what the AI said it does, and you need to check this via a traditional lit search. And if the AI says that something is not in the literature, you cannot trust it. But when this works, this speeds things up a lot.