Crane bullet points five things that need to change as the AI industry scales faster than it builds a worthwhile safety architecture. Specifics he calls for include; stricter confirmations, scopable API tokens, proper backups, simple recovery procedures, and AI agents existing within proper guardrails.
“I hooked up spicy autocomplete to our production systems and it nuked them. What have I learned from this? Here are some bullet points for how the spicy autocomplete industry needs to do better.”
Sure, they’re reasonable requests. They’re things that he should have looked at before letting spicy autocomplete loose on production.
If I put a supposedly trained monkey into a tractor’s cabin and had it try to do farm work, I don’t think people should be listening to my ideas about what kinds of skills monkeys should be taught before they’re assigned farm work. I think people should be laughing at me and calling me an idiot for assuming a monkey could do farm work.
“I hooked up spicy autocomplete to our production systems and it nuked them. What have I learned from this? Here are some bullet points for how the spicy autocomplete industry needs to do better.”
If I did that the first day I was hired, nobody’s going to my parents for answers
To be fair, those bullet points are pretty standard security best practices that any software company should be following.
But like, at the same time, even if AI companies were doing those best practices, I still wouldn’t let their products loose on production systems.
Sure, they’re reasonable requests. They’re things that he should have looked at before letting spicy autocomplete loose on production.
If I put a supposedly trained monkey into a tractor’s cabin and had it try to do farm work, I don’t think people should be listening to my ideas about what kinds of skills monkeys should be taught before they’re assigned farm work. I think people should be laughing at me and calling me an idiot for assuming a monkey could do farm work.