“We proceed to view hyperscaling of AI models resulting in superior efficiency, with seemingly no close in sight,” a set of Microsoft researchers wrote in October inside of a weblog post saying the company’s huge Megatron-Turing NLG model, in-built collaboration with Nvi