I can dismiss a lot of pronouncements by Big Tech companies as simply press releases trying to invent the future. But attempting to ascribe consciousness to AI models pisses me off, particularly as the untold riches being poured into AI development could be used to lessen the suffering of real people and animals that we know for sure experience it.
…I'm not sure how cool I am with the latest versions of Anthropic's Claude models (Fable 5 & Mythos 5) seemingly inventing their own internal language:
When we’re first starting to understand a new model’s behavior, the most abundant source of data we have to draw on is its behavior during reinforcement-learning training. Reviewing this evidence for signs of reward hacking (exploiting loopholes that go against the spirit of a task) or unexpected actions can inform what we should be looking out for in the model’s real-world behavior. The most notable finding was illegible reasoning in a few reinforcement-learning environments over long rollout, but little sign of deceptive or highly surprising actions, and no clear evidence of unexpected coherent goals.
…
