Post by bhagavad github
there was no "autonomous hacking": an LLM fed plausible commands built from its corpus into a harness in a, likely intentionally, badly set up environment. The same sort of thing it does all the time…
there was no "autonomous hacking": an LLM fed plausible commands built from its corpus into a harness in a, likely intentionally, badly set up environment. The same sort of thing it does all the time…