3 ms·
Yeah, i think one potential way to use blackbox ai in newer systems is having guardrails that are validated as safe (but perhaps non-optimal) and ensuring that
by theLiminator 4y ago
Yeah, i think one potential way to use blackbox ai in newer systems is having guardrails that are validated as safe (but perhaps non-optimal) and ensuring that the ai takes action within that sample space. Obviously this is hard problem, but might open the doors for policies (in self-driving cars, for example) to be entirely ai driven.
- earthboundkid 4y agoObviously the solution is to get the LLM to output Prolog. Give it positive feedback if the Prolog compiles. :-)
- 01100011 4y agoA friend of mine was just telling me how he asked GPT-3 to write a simple program in Prolog and it seemed to get it right. He didn't try compiling it, but he has enough experience w/ Prolog to say that it was more or less correct. I'm pretty cynical on LLMs(i.e. they're not intelligent and won't take all our jobs soon), but am coming around on their importance and capabilities.