Microsoft is developing MAI-1, an artificial intelligence aimed at competing with Google and OpenAI. A recent report reveals that the tech giant is training its own large language model (LLM) to rival GPT-4, Claude, and other powerful AIs. This initiative marks Microsoft’s first major endeavor since its multi-billion dollar investment in OpenAI.
Please follow us on Facebook and Twitter.
As reported by The Information, Microsoft is spearheading the development of a new artificial intelligence model under the guidance of Mustafá Sulaymán. Sulaymán, the co-founder of DeepMind and the newly appointed director of Microsoft AI, leads a groundbreaking project that deviates from previous experiments. Dubbed MAI-1, this model boasts 500 billion parameters, half the size of GPT-4, and offers adaptability during training.
Internal sources reveal that MAI-1 is Microsoft’s proprietary model trained within its premises. Leveraging training data and technologies from Inflection AI, a company co-founded by Sulaymán, Microsoft embarked on this endeavor. Following Sulaymán’s recruitment in March, Microsoft acquired Inflection AI along with its entire workforce, agreeing to a hefty $650 million licensing fee for its AI software.
MAI-1 inherits training data and technologies from Inflection AI to carve out its unique language model. This strategic move not only mitigates Microsoft’s reliance on OpenAI but also facilitates the development of AI tailored to address specific challenges within its services.
Microsoft Gets Fully into Artificial Intelligence with MAI-1
Microsoft is heavily banking on artificial intelligence, evident from its significant investments in OpenAI and the integration of AI across its services. Unlike some industries like mobile phones, Microsoft has shown swift responsiveness.
Before introducing MAI-1, Microsoft unveiled Phi-3 Mini, an AI capable of matching GP-3.5’s performance at a fraction of the cost. Phi-3 Mini boasts around 3.8 billion parameters and was trained using children’s stories. Microsoft claims it learns similarly to how children do when engaged in educational games.
Microsoft aims to elevate its AI endeavors to a profitable level, envisioning MAI-1 as a lucrative investment. However, managing 500 billion parameters demands substantial processing power, akin to Meta’s requirements. Redmond executives may need to engage with Uncle Jensen to secure the necessary hardware.
Details regarding MAI-1 and Microsoft’s other AI initiatives remain undisclosed for now. However, the company is anticipated to unveil updates during a special event on May 20. While the event’s main highlight is the Surface with Snapdragon X Elite, Microsoft will also shed light on its near-term AI strategy.
Read Also: Reportedly, Kaspersky’s AI Used in Russian Military Drones