1 Comment
⭠ Return to thread

what does “compute” matter when we’ve run out of training data, and the models have the same theoretical holes, such as not being able to distinguish correlation from causation, that they have always have?

Expand full comment