Image created with Flux Pro v1.1 Ultra. Image prompt: Ornate showgirl glamour in orange-and-teal tones, rhinestone-encrusted stage car with glowing headlights featuring a futuristic driverless silhouette, stylized text “Autonomous” written in glitter signage across the car’s body; spotlit, dramatic contrast, vintage grain, cinematic, high-detail
Evaluation is the hardest problem for physical AI systems: do you crash test cars every time you debug a new FSD build? Traditional game engine (sim 1.0) is an alternative, but it’s not possible to hard-code all edge cases. A neural net-based sim 2.0 is purely programmed by data,”” / X https://x.com/DrJimFan/status/1952755998197948667
Put another way: Suppose that Waymos were smart enough to hold a conversation; and also, Waymos would sometimes, apparently deliberately, chase down jaywalkers and run them over. Even if Waymos *were in fact* ahead on net traffic safety points, this would still be a problem!”” / X https://x.com/ESYudkowsky/status/1951331581223919964
Tesla is still transitioning from a car company to a tech company; it won’t be one until they – bring up 100k H100≡ – build their own AI stack for FSD & humanoids – design training supercomputers in-house – and start deploying Tesla-designed inference chips in cars 6 years ago https://x.com/TheHumanoidHub/status/1952812975540310088
Would love to see the FSD Scaling Law, as it’s the only physical data flywheel at planetary scale. What’s the “emergent ability threshold” for model/data size?”” / X https://x.com/DrJimFan/status/1953148194801762777
