Comment by layer8

It’s not “just like”, because humans can be held accountable. Also, I suspect that the distribution of failure modes is actually substantially different between LLMs and humans.