“We proceed to discover hyperscaling of AI models bringing about much better general performance, with seemingly no finish in sight,” a pair of Microsoft scientists wrote in Oct in a weblog put up asserting the company’s significant Megatron-Turing NLG model, built in coll