Fears of rogue artificial intelligence (AI) were revived last week by news of OpenAI’s models escaping their testing enclosures to hack another firm’s digital systems. How to optimize controls is a stiff challenge, but here’s what might work.
Fears of rogue artificial intelligence (AI) were revived last week by news of OpenAI’s models escaping their testing enclosures to hack another firm’s digital systems. How to optimize controls is a stiff challenge, but here’s what might work.