A report published on Monday revealed that Microsoft is working on an in-house AI language model that’s internally called “MAI-1”.
The project is currently being overseen by Mustafa Suleyman—a recent hire who co-founded Google DeepMind and was the CEO of AI startup Inflection.
This will be the first large-scale AI model directly developed by Microsoft.
In a report published on Monday (May 6), two Microsoft employees revealed that the company is planning to make a mark in the AI industry with its own AI language model that would be large enough to compete with Google and OpenAI.
The AI model has no official name yet. But internally, it’s called the “MAI-1”.
The project is being overseen by Mustafa Suleyman who was recently hired by Microsoft. Suleyman previously served as the CEO of Inflection (an AI startup) and was also a co-founder of Google’s DeepMind. So, it’s safe to say that he’s probably the best man for the job.
One of the most important points to note is that Microsoft has previously only worked on smaller, open-source AI tools. For example, just last month it launched a small AI tool Phi-3-mini that aimed to attract a larger customer base with its significantly affordable pricing.
However, Microsoft has never directly invested in anything at a larger scale—even Phi-3-mini is meant for devices with limited computational power, such as laptops and smartphones. MAI-1 will be its first attempt at a large language model (LLM), meaning it would also be more expensive.
To give you an idea of how large this model will be, it will consist of a whopping, potentially class-leading 500 billion parameters—GPT-3 has 175 billion parameters and Phi-3-mini just 3.8 billion parameters.
Note: Parameters are variables in AI models whose values are adjusted during the training period to help the tool understand what should be the desired output for a given input.
We don’t have a ton of detail about how far along the MAI-1 project is, but the report revealed that the company has been setting aside a number of servers with Nvidia’s GPUs, along with bulk data, for training its new AI model.
The report also said that although MAI-1 is not copied from AI firm Inflection, it’s certainly based on some training data acquired by that startup.
The exact purpose of this AI model is yet to be decided—it will probably depend on how efficient it is.
Is This a Good Time to Invest in an AI Model?
Whether launching a new AI model at this moment is a good idea or not remains a big question.
For starters, tools like ChatGPT and Gemini are already dominating the industry.
Secondly, finding enough data to train these models is becoming a big challenge, as companies can’t just pick content from the internet.
Just recently, both OpenAI and Microsoft were sued by 8 US newspapers for using their articles without consent or compensation. Similar lawsuits have also been filed against these companies in the past, most notably by the New York Times.
Last but not least, Microsoft should also expect a fair amount of competition from xAI chatbot GrokAI, which was recently made open-source.
However, the fact remains that these are all speculations from information revealed by inside sources—what Microsoft has really planned for its new AI model is still a mystery. Stay tuned for more information.
The Tech Report editorial policy is centered on providing helpful, accurate content that offers real value to our readers. We only work with experienced writers who have specific knowledge in the topics they cover, including latest developments in technology, online privacy, cryptocurrencies, software, and more. Our editorial policy ensures that each topic is researched and curated by our in-house editors. We maintain rigorous journalistic standards, and every article is 100% written by real authors.
>>> Read full article>>>
Copyright for syndicated content belongs to the linked Source : TechReport – https://techreport.com/news/microsoft-ai-language-model/