X @The Economist

“Trying to teach AI systems only to follow good commands is not going well.” Discover why the security flaw in large language models is hard to fix on “The Intelligence” https://t.co/nmQO68i5ei https://t.co/yJdyrUH4If ...