Even if we get ultra-capable frontier models, we’ll need lots of GPUs to run them at scale. Currently, our installed GPU stock isn’t enough to automate the white-collar workforce or cause extinction. We need more!
Tyler Cowen thinks we should predict AI progress using asset prices. He’s usually thinking about interest rates, but what about NVIDIA’s stock price? Since NVIDIA makes almost all of the GPUs used to train and run AI models, and since GPUs generate almost all of NVIDIA’s sales, NVIDIA’s market capitalization basically represents the discounted value of future cash flows from the GPU market.
Using a reverse DCF model, you can estimate how much NVIDIA must grow to generate cash... (read 881 more words →)
This critique seemed very persuasive to me. Thank you for putting it together.
The timeline forecast is blended distribution of the superexponential (40% - 45%), exponential (45% - 50%), and subexponential (10%). I would think there is going to be a pretty consistent rank-ordering, where almost all of the mass of the superexponential is earlier than the almost all of the mass of the exponential. Similarly, almost all of the mass of the subexponential is going to be later than either the exponential or superexponential.
This is a simplification, but running with it for a moment... Because the super-exponential block contains < 50 % of the total probability mass, the overall median will come... (read more)