[Terr_]

> There will be demand for gpt's trained on an actual engineering material and it could actually be a huge gamechanger for that market.

I imagine there will also be a lot of kinda-fraudulent supply from people who think: "I'll just take a cheap/commodity (badly) trained LLM, find just the right set of whack-a-mole prompts to make it appear to be making good output, and until customers catch-on the difference is pure profit."

Or perhaps they're open about it, and many customers just decide bad results cheap is better than premium data, which is... not a heartening thought.