“We keep on to check out hyperscaling of AI models leading to greater overall performance, with seemingly no conclude in sight,” a set of Microsoft scientists wrote in Oct inside of a weblog put up saying the company’s significant Megatron-Turing NLG model, built in collab