- Microsoft is building its own AI model called MAI-1, The Information reported.
- The language model will be separate from OpenAI's GPT-4 and will be overseen by Mustafa Suleyman.
- This signals that Microsoft is prepared to reduce its dependence on ChatGPT makers as the AI wars intensify.
Microsoft is reportedly working on its own AI model separately from OpenAI. This is CEO Satya Nadella's chance to prove that her company doesn't need a ChatGPT maker to gain an edge in the AI wars.
Microsoft is working on building an in-house large-scale language model called MAI-1, The Information reported earlier this week. This is the first standalone AI the software giant is building since the company poured $10 billion into OpenAI for the rights to power generative AI tools such as Copilot with GPT-4, the basis for ChatGPT. It's a model.
The creation of MAI-1 could help address some concerns that Microsoft is too focused on its ongoing OpenAI partnership as the race to develop the best generative AI tools intensifies.
Regarding OpenAI, CEO Nadella once said that Microsoft is “beneath them, above them, and around them,” but one investor pointed out that this was a statement made by a “hotshot.”
The internal project is being overseen by Mustafa Suleiman, co-founder of Google's DeepMind. He joined Microsoft in March to lead its AI division after leaving his startup, Inflection AI, as CEO.
To train MAI-1, Microsoft is said to be compiling a dataset that includes external sources such as text from ChatGPT and public information on the web, a source told The Information. Some of the data used to train Inflection's models may also be used to train Microsoft's new internal models, two Microsoft employees told the publication.
MA-1 is expected to be larger than open-source models created by rivals such as Meta and Mistral, and will require significant amounts of computing power and training data. This is in addition to the smaller, less advanced models that Microsoft is developing for its smartphone apps, reflecting the tech giant's multifaceted approach to developing advanced AI.
After news of the project broke, Microsoft CTO Kevin Scott said the relationship between Microsoft and OpenAI remains symbiotic, with both companies being key drivers in building the best AI. .
“We are building increasingly powerful supercomputers for OpenAI to train models that will set the pace for the entire field, and we will continue on this path well into the future,” Scott said in a report in The Information. I wrote this in a post on LinkedIn in response to this. “There is no end in sight to the growing impact of our collaboration.”
However, it's not yet clear what the internal model will look like, and Microsoft did not immediately respond to Business Insider's request for comment for further details.
More information about MAI-1 is likely to be revealed at Microsoft's Build conference for developers in Seattle from May 21-23.
Axel Springer, Business Insider's parent company, has a global deal that allows OpenAI to train models based on its media brands' reporting.
