Google DeepMind CEO Demis Hassabis:
— We don’t know how to contain an AGI
— We don’t know how to test for dangerous AI capabilities
— We should cooperate internationally and build an AI CERN to ensure AI is developed safely pic.twitter.com/cafd4Nxwiq— ControlAI (@ai_ctrl) August 15, 2024
Demis Hassabis on the risks of open-source AI:
— In 2-4 years, if powerful AI is misused, it could cause serious damage.
— With open-source AI, if something goes wrong and a bad actor misuses it, you can’t recall it. There’s no pulling it back. pic.twitter.com/nFe8EX6nq7— ControlAI (@ai_ctrl) August 16, 2024