“We go on to view hyperscaling of AI models bringing about far better performance, with seemingly no stop in sight,” a set of Microsoft scientists wrote in October in the web site write-up announcing the company’s large Megatron-Turing NLG model, built in collaboration wit