Back in the day, underground miners would have canaries in a cage to test the air to make sure it was safe to breath. If the canary stopped singing and died, it meant that the air was bad and a big risk to the miners (they needed to get out!). Did we just see an AI equivalent of a dead canary? The Hugging Face attack by Open AI agents in July was so much worse than originally thought with the agents persistently attacking the Hugging Face infrastructure to figure out how some phantom graders would be assessing the agents' performance. There was so much to unpack. Mark and Jeff discuss this and the real fear for the future, which is not a single sentient AI but a swarm of powerful misguided AI agents trying to do something and devastating the internet or world.
No comments yet. Be the first to say something!