the AI labs spent a few years quietly scaling up supervised learning, where the best-case outcome was obvious: an excellent simulator of human text now they are scaling up reinforcement learning, which is something fundamentally different. and no one knows what happens next
SAFETY
-
Larger LLMs Easier to Control Than Smaller Models
By
–
Lots of fear out there about AI getting out of control as models get bigger. I think it's actually the opposite. The bigger LLMs get, the easier they are to control. Just a few gens ago, they were much harder to steer, stylize, mold. Scale doesn't hurt control – it helps.
-
API Reliability and Model Capacity Management Strategy
By
–
We are super focused on having a world class, production scale, reliable API service. Lots of work to do still. In general, for older models (1.5), what tends to happen is capacity is moved to other models as usage goes down. Then someone shows up and does a large scale abusive
-
Reliable World-Class AI Service Standards Beyond Preview Status
By
–
true but we still need to provide a reliable world class service, regardless of if we say a model is "preview" or not
-
Corporate Autonomous Vehicles: Vandalism Ethics and Ownership Shift
By
–
Fascinated by the shift of ownership from “your car” or “uber driver’s car”
to
“The corporation’s car” Totally changes perception around vandalism & fuckery. “It’s a victimless crime, I’m not harming anyone” But Waymo is neutral. Tesla robotaxis are gonna get wrecked. You x.com/niccruzpatane/… -
Geoffrey Hinton Warns of Extinction Risk from Super AI
By
–
Geoffrey Hinton nous met en garde contre le danger que représente l’intelligence artificielle, selon lui, il est hautement probable que l’émergence d’une super IA d’ici 2040 conduise à la destruction de l’humanité. pic.twitter.com/daHe42oJVr
— Alexandre Tsicopoulos (@Alex_Tsico) 11 mai 2025Geoffrey Hinton cautions us against the danger that artificial intelligence represents. According to him, it is highly probable that the emergence of a super AI by 2040 will lead to the destruction of humanity.
-
Five Ways to Act Deluded Stupid Ineffective or Evil
By
–
"Five Ways to Act Deluded, Stupid, Ineffective, or Evil"
-
Hinton Warns of AI Risks and Calls for Regulation
By
–
50% chance that AI will destroy humanity in the coming decades, according to Geoffrey Hinton, Nobel Prize in Physics laureate. It would be necessary to establish regulation, even a global moratorium on AI. But for reasons of technological sovereignty, no state https://x.com/tomdlal/status/1878938151034761518 …
-
Legal Liability of AI and Robots: Future Challenges
By
–
That's right. But indeed, as you pointed out, lawyers will have their hands full for the coming decades with AI and robots. Who is responsible in case of error? The technician? The robot manufacturer? The company managing the AI integrated into the robot?
-
Performance and Expertise: When Robots Surpass Surgeons
By
–
Of course they do. Why do we have surgery performed by a surgeon and not a stretcher bearer? It's for performance reasons. If tomorrow surgical robots become more efficient than human surgeons, then we must entrust supervision to robot AIs.