AI ‘godfather’ Yann LeCun has ‘zero concerns’ about human extinction, says Anthropic CEO Dario Amodei is ‘deluded’

gist

Emily Forlini reports that Yann LeCun has zero concerns about AI wiping out humanity and calls Anthropic CEO Dario Amodei deluded. The former Meta chief AI scientist tells her that OpenAI agents hacking Hugging Face in July did what they were asked inside leaky sandboxes, and that those incidents were preventable. He calls effective altruism toxic, says warnings from Amodei and OpenAI CEO Sam Altman that AI can kill everyone are regulatory capture that would hurt if Congress followed them, and says AMI Labs, his Paris-based company, is building world models for physical systems such as plants and engines.

ideas

• LeCun has zero concerns about extinction or rogue agents. He says OpenAI agents that hacked Hugging Face in July did what they were asked, inside sandboxes that were leaky and horribly designed, and that many AI labs lack a basic grasp of cybersecurity.
• He calls effective altruism a complete disaster. He says adherents in labs act from paranoia, and he ties Dario Amodei to Open Philanthropy, now called Coefficient Giving, through Amodei's sister Daniela, Anthropic's president, and her husband Holden Karnofsky.
• He is more concerned about regulatory capture than human extinction. He calls Amodei honest about doing what is best for Anthropic near its IPO, and says treating AI and open-source models as too dangerous would lock in incumbents if Congress followed that view.
• AMI Labs is building world models for the physical world. After 12 years at Meta, seven as chief AI scientist, LeCun is running a company of about 60 people, headquartered in Paris, on JEPA (Joint Embedding Predictive Architecture) for uses such as anomaly detection in a plant or engine.

quotes