Effectively no one was but LLM's are precisely "ML predictive algorithms". That neural networks more broadly would scale at all on gaming chips is plenty foresight to be impressed with.