Microsoft revises agreement with OpenAI, targeting the new track of "superintelligence"

Microsoft is developing a more powerful form of artificial intelligence called "superintelligence," hoping to achieve advancements in fields such as medicine and materials science.

On November 6, Mustafa Suleyman, head of Microsoft's AI department, published a blog post announcing that he will lead the newly established "Humanitarian Superintelligence" team, which will aim for a more ambitious hypothetical milestone than general artificial intelligence (AGI). Previously, Microsoft had reached an agreement with OpenAI not to participate in the development of AGI.

In the blog post released on Thursday, Suleyman stated that he would promote the development of personal AI assistants and focus on breakthroughs in healthcare and clean energy. He said:

"If general artificial intelligence (AGI) is generally seen as the threshold at which AI can perform at human levels across all tasks, then superintelligence is the stage at which it can far exceed human performance."

This announcement comes as Microsoft and OpenAI have just reached a renegotiated agreement. The new agreement defines Microsoft's equity stake in the startup and modifies some terms of the relationship between the two parties, including lifting the previous ban on Microsoft developing advanced AI tools. This ban had restricted the Redmond, Washington-based company to developing smaller, less powerful models, while the models used by ChatGPT are significantly more powerful.

Thursday's announcement effectively formalizes a project that Microsoft has been preparing since March of last year. At that time, Microsoft hired Suleyman and obtained intellectual property licenses from his startup Inflection AI. By restructuring the Microsoft team and recruiting new employees, Suleyman began building a new series of Microsoft AI models. However, so far, these models are still significantly smaller in scale than the strongest products from OpenAI or Alphabet's Google