nostr:npub1zl93dtqgvm78n7gqsse3tncf3x958p9dnyxr036y3qqsy0c7gq3swg76w7 

[#generativeAI is] emitting text that looks like what you want. 

That exactly the point. You get it.

Nobody who can be liable for copyright infringement or needs to know the origin of each fact in a statement that will be published or used in decisions can use something like #chatGPT. It is transparent as lead as to where it got any fact.

What I understand is that people or entities that can be liable use is curated and cleared models that give a good illusion of understanding the language so it can parse intent and sentiment into statistics.

Since the generative AI will try to emit text that looks like what you want, there's the concept of "prompt tuning" going hand and hand with the "curated data" it can access. Prompts that provide correct answers most of the time are "canned."  Programs that follow these rule will only use the canned prompts.

This is a generalization of the process presenters talked about.

Practically everyone thinks #chatGPT is a monster, but many point out it will spur lawmakers to get copyright law and regulation updated quicker, lest the courts decide such things haphazardly and unfairly.