“We continue to discover hyperscaling of AI models leading to better general performance, with seemingly no conclude in sight,” a pair of Microsoft scientists wrote in October in a very website write-up saying the company’s massive Megatron-Turing NLG model, in-built colla