Microsoft is increasing its AI footprint with the of two new fashions that its groups educated utterly in-house. MAI-Voice-1 is the tech main’s first pure speech era mannequin, whereas MAI-1-preview is text-based and is the corporate’s first basis mannequin educated end-to-end. MAI-Voice-1 is presently getting used within the Copilot Every day and Podcast options. Microsoft has made MAI-1-preview obtainable for public checks on LMArena, and can start previewing it in choose Copilot conditions within the coming weeks.
In an interview with , Microsoft AI division chief Mustafa Suleyman stated the pair of fashions was developed with a give attention to effectivity and cost-effectiveness. MAI-Voice-1 runs on a single GPU and MAI-1-preview was educated on about 15,000 Nvidia H-100 GPUs. For context, different fashions, akin to xAI’s Grok, took greater than 100,000 of these chips for coaching. “More and more, the artwork and craft of coaching fashions is deciding on the proper knowledge and never losing any of your flops on pointless tokens that didn’t really train your mannequin very a lot,” Suleyman stated.
Though it’s getting used to check the in-house fashions, Microsoft Copilot is primarily constructed on OpenAI’s GPT tech. The choice to construct its personal fashions, regardless of having sunk within the newer AI firm, signifies that Microsoft needs to be an impartial competitor on this house. Whereas that would take time to succeed in parity with the businesses which have emerged as forerunners in AI growth, Suleyman advised Semafor that Microsoft has “an unlimited five-year roadmap that we’re investing in quarter after quarter.” With some issues arising that AI might be going through a bubble-pop, Microsoft’s timeline will should be aggressive to make sure taking the impartial path is worth it.
