Microsoft Unveils Plans for Advanced AI Model to Rival Google, OpenAI
The report said that Microsoft is training a new in-house AI language system large enough to compete with Alphabet, Google, and OpenAI. The exact purpose of the model is yet to be determined, and it depends on its efficiency.
The report suggests that Microsoft will likely examine the new model as soon as it holds its build developer meeting later this month. When contacted by Reuters, Microsoft declined to comment.
The MAI-1 will be "significantly larger" than previous smaller, open-source models Microsoft previously trained on, which means it will be more expensive, the report said.
Microsoft launched a smaller version of artificial intelligence called the Phi-3-mini last month, and it looks set to appeal to a broader consumer base with more cost-effective options.
The report said Microsoft isolated a large group of servers equipped with Nvidia graphic processing units to improve the model with more data.
MAI-1 will have about 500 billion parameters, the report said. The OpenAI GPT-4 reportedly has one billion parameters, and the Phi-3 mini will measure 3.8 billion parameters.
In March, Microsoft tapped Suleiman to head its revamped consumer AI model unit and hired more Inflection employees.
The new model cannot run from Inflection, although it can build on training data from startups, the report added.
Conclusion: As Microsoft embarks on this ambitious Artificial Intelligence journey, the tech community is eagerly awaiting the unveiling of its groundbreaking AI infrastructure and transformative impact on society. With innovation as a guiding principle, Microsoft is committed to progress and unlocking the full potential of AI beyond intelligence and the inclusiveness it can create in the universe.