“We continue to discover hyperscaling of AI models leading to greater performance, with seemingly no conclude in sight,” a set of Microsoft researchers wrote in Oct in a site publish asserting the company’s huge Megatron-Turing NLG model, in-built collaboration with Nvidia