“We carry on to find out hyperscaling of AI models resulting in greater general performance, with seemingly no end in sight,” a pair of Microsoft researchers wrote in October inside a weblog submit saying the company’s enormous Megatron-Turing NLG model, inbuilt collaborat