“We keep on to find out hyperscaling of AI models bringing about improved efficiency, with seemingly no conclusion in sight,” a pair of Microsoft researchers wrote in October in the blog site write-up announcing the company’s massive Megatron-Turing NLG model, inbuilt coll