@PatrickHeizer The margin of error in the system is basically gone. One significant event will likely tip the domino’s. My timeline is less than 1 year.
@beffjezos When are people going to admit benchmarks are just ridiculous. They are not a prediction of how good a model can be. It’s like a bloody university exam.
@_sholtodouglas Too much weight is placed on benchmarks. They don't necessarily indicate if a model is good for a real-world task. They are like exams which do not predict real world performance.