“We carry on to find out hyperscaling of AI models resulting in much better performance, with seemingly no finish in sight,” a set of Microsoft scientists wrote in Oct within a blog submit announcing the company’s enormous Megatron-Turing NLG model, built in collaboration