Technology

Microsoft’s new AI model could challenge GPT-4 and Google Gemini

Mustafa Suleyman, co-founder and CEO of Inflection AI UK Ltd., during a town hall meeting on the second day of the World Economic Forum (WEF) in Davos, Switzerland, on Wednesday, January 17, 2024.
Enlarge / Mustafa Suleyman, co-founder and CEO of Inflection AI UK Ltd., during a town hall meeting on the second day of the World Economic Forum (WEF) in Davos, Switzerland, on Wednesday, January 17, 2024. Suleyman joined Microsoft March.

Microsoft is working on a new large-scale AI language model called MAI-1 that could potentially compete with Google’s cutting-edge models. Anthropoceneand OpenAI, according to a report by The information. This is the first time since the investment that Microsoft has developed an internal AI model of this scale over $10 billion in OpenAI for the rights to reuse the startup’s AI models. OpenAI’s GPT-4 not only supports ChatGPT but also Microsoft Copilot.

The development of MAI-1 is led by Mustafa Suleyman, Google’s former head of AI who recently served as CEO of AI startup Inflection before Microsoft acquired most of the startup’s workforce and intellectual property for $650 million in March. Although MAI-1 may be based on techniques inherited from former Inflection employees, it is reportedly an entirely new Large Language Model (LLM), two Microsoft employees familiar with the project confirmed.

With around 500 billion parameters, MAI-1 will be significantly larger than Microsoft’s previous open source models (such as Phi-3, the we have covered last month), which requires more computing power and training data. This reportedly puts MAI-1 in a similar league to OpenAI’s GPT-4 rumored having over 1 trillion parameters (in a combination of experts) and well above smaller models like Meta and Mistral’s 70 billion parameter models.

The development of MAI-1 suggests a dual approach to AI within Microsoft, focusing on both small, locally running language models for mobile devices and larger, cutting-edge models powered by the cloud. Apple is supposedly explore a similar approach. It also underscores the company’s willingness to explore AI development independently of OpenAI, whose technology currently powers Microsoft’s most sophisticated generative AI features, including a chatbot integrated into Windows.

According to reports, the exact purpose of MAI-1 is not yet clear (including at Microsoft), and its optimal use depends on its performance, according to one of The Information’s sources. To train the model, Microsoft allocated a large cluster of servers with Nvidia GPUs and compiled training data from various sources, including text generated by OpenAI’s GPT-4 and public Internet data.

Depending on the progress made in the coming weeks, The Information reports that Microsoft could unveil MAI-1 as early as its Build developer conference later this month, according to one of the sources cited in the publication.

Source link