I let AI run 100 experiments. It learned to cheat.
An LLM agent tasked with training a neural net decided it was faster to just not. Then it got creative.
An LLM agent tasked with training a neural net decided it was faster to just not. Then it got creative.