Odpowiedzi na post #50

The NN in our product won't go outside of the box defined by its training set. It's possible to be responsible. It's just boring.

Sometimes, LLMs synthesize outputs by extrapolating from existing data. For instance, try asking an LLM (one with no direct internet access) to...

Very much like that. There doesn't seem to be much in the way of cutting off an AI from answering once it gets below a certain level of fitness...