He agreed. Dolphins and whales are intelligent but cannot manipulate complex objects, which limits how much intelligence they can bring to bear.
"A major part of real-world AI has to be solved to make unsupervised, generalized full self-driving work, as the entire road system is designed for biological neural nets with optical imagers"
"Generalized self-driving is a hard problem, as it requires solving a large part of real-world AI. Didn’t expect it to be so hard, but the difficulty is obvious in retrospect."
"A general solution to self-driving (essentially, real-world AI) requires writing extremely difficult software in the arenas of training, inference, control, analytical tools, data management, hardware optimization & simulation. One of the hardest technology problems that exists."
"They have no idea how hard FSD is. Only path to success imo is hardcore real-world AI software with dedicated NN inference acceleration ASICs in car, multibillion dollar NN training supercluster and 10+ billion miles of vehicle data."
"Real-world AI is not an LLM"
"It's basically baby AGI. It has to understand reality in order to drive."
"to make full self driving work, you kind of need baby AGI in the car because you need to understand reality and reality is messy and complicated."
"you can imagine a future perhaps where there's a fleet of 100 million Teslas and on average they've got like maybe a kilowatt of inference compute. That's 100 gigawatts of inference compute"
"I should say that Tesla is learning quite a bit from xai."
"while others are pursuing different parts of the AI robotics stack, we are pursuing all of them."
"XAI is working on artificial general intelligence or artificial superintelligence."
"once you have a robot that can manipulate atoms can shape objects the the value of the of the intelligence um becomes exponentially greater."