Microsoft seems to be training a new, in-house AI language model large enough to rival with those developed by Alphabet’s Google and OpenAI, The Information reports. The new model, internally codenamed MAI-1, is being overseen by Mustafa Suleyman, who is the co-founder of Google Deepmind and ex-CEO of Inflection AI startup. The report cites two Microsoft employees with knowledge of the effort.
The exact purpose behind developing the model is still unclear and it will depend on the performance of the model. Microsoft could showcase the new model during its Build developer conference later this month, the report suggests.
MAI-1 will be “far larger” than the previous smaller, open-source models that the company has trained before. The report also suggests this will make the new model more expensive.
Also Read: Apple To Give Siri For iOS 18 Major Upgrade Via Apple’s Ajax LLM
Will Microsoft’s New AI Model Take Training Data Form Inflection?
Last month, Microsoft launched a smaller AI model dubbed Phi-3-mini as it seems to attract a bigger client base with cost-effective alternatives. The company seems to have invested billions of dollars in OpenAI and has deployed the ChatGPT maker’s tech across its suite of productivity software, enabling it to gain a huge lead in the AI race.
Microsoft has also set up a huge cluster of servers fitted with NVIDIA’s GPUs along with large amounts of data to enhance the model, as per the report. MAI-1 will have around 500 billion parameters, the report suggested, while OpenAI’s GPT-4 is reported to have one trillion parameters and Phi-3-mini has 3.8 billion parameters.
Microsoft brought in Suleyman in March as the head of the new AI unit and hired many employees of Inflection. The new model is not being developed at Inflection; however, it could build training data from the startup, the report suggests.
Also Read: Engineers Report Burnout As Tech Companies Chase AI Gold Rush
Microsoft Dragged To Court Over Copyright Infringement
Many news organizations, including the New York Daily News, Orlando Sentinel, Chicago Tribune, San Jose Mercury News, and four more firms have filed a lawsuit against Microsoft and OpenAI over copyright infringement allegations.
The hedge fund Alden Global Capital owns all of these publications, and the lawsuit claims that both OpenAI and Microsoft trained using their content without any compensation or consent.
The plaintiffs also showed the evidence including several excerpts from conversations with ChatGPT and Copilot. The evidence showed that both AI chatbots recreated lengthy excerpts of specific articles on prompt, showcasing that their training datasets included the content from the articles.