Post #3991233

@pludikovsky@chaos.social You may want to learn more about LLMs and their limitations if you want to create scenarios that reflect the reality of what is possible. I'm not saying that training data wouldn't be helpful to detect these issues - just that the LLM seems like the wrong tool to actually solve the problem - it seems more like the tool we have rather than the one we need. Ultimately, your proposal would end up where are now - the tools don't do what you want.