Raw LLM Responses

Our current approach for AI will never be appropriate for engineering. LLMs are bullshit generators. They are only designed to produce a response that sounds like a person made it. They have no concept of truth or understanding. At best LLM AI would produce results that sounded credible. This is a terrible thing for engineering. You need results that make logical sense and have been thought through, not results that “sound” right. There is already a big problem of humans copying engineering work and misapplying it to situations where it doesn’t apply, and humans are capable of knowing better. AI based on language models can’t do this as it has no way to know what is right or wrong. That’s not to say people won’t use AI for engineering. Just that they shouldn’t, nor should anyone trust the work of such a program. It would be like taking away the library of conversations the LLMs use to fake being cognizant and expecting it to converse based on an understanding of the meaning of the actual words. The AI doesn’t understand words and it never did, that’s not how it works.