Academic Daniel Litt says the expense of generating one high-quality AI result is lower than the cost of equivalent output from a human expert even after including failed attempts
— Yacine highlighted a thread claiming AI models exceed human costs in compute time