Following Meta, Microsoft (MS) has also begun research and development on "superintelligence" that surpasses humans. Mustafa Suleyman, CEO of the MS AI Group (MAI), will lead the team.
On the 6th (local time), CEO Suleyman said on the official blog, "We have established the MAI superintelligence team that I lead," and "I hope this becomes the world's best place for AI research and development."
Artificial general intelligence (AGI), widely known through OpenAI, refers to AI that has a human-like level across every institutional sector. But superintelligence refers to AI that goes beyond that level.
MS released a full-fledged superintelligence development plan because, during OpenAI's recent governance overhaul to convert into a public corporation (PBC), MS became able to develop its own AI. Earlier in 2019, when MS invested in OpenAI, it signed a contract under which it would hold rights to AI models developed by OpenAI but would not use them to develop AGI.
However, during the recent governance overhaul at OpenAI, the relevant clause was revised, allowing MS to pursue independent AI development.
Accordingly, MS rehired Suleyman, the DeepMind co-founder who left the company in 2022, in Mar. last year to build the "MAI" organization.
In his announcement that day, CEO Suleyman said, "MS AI aims for 'human-centered superintelligence' (HSI)."
He defined HSI as "astonishingly advanced AI capabilities that always operate in a way that serves people and all of humanity." It is not an unlimited entity with a high level of autonomy, but AI that is carefully tuned and operates within a context-appropriate, limited scope.
He added that his team's mission is "not to pursue directionless technical goals or hollow challenges, nor to climb a mountain just because it is there, but to solve concrete real-world problems in a way grounded in reality and subject to control," and emphasized, "We reject the narrative of an AGI race."
As examples of HSI, he cited "AI companions" that help humans, "medical superintelligence," and "clean energy AI."
Earlier, in Jul., Meta also established the "Meta Superintelligence Lab" (MSL). At the time, as Meta fell behind in AI competition with OpenAI and Google, it reportedly set a goal of skipping steps to develop superintelligence.
To that end, Meta CEO Mark Zuckerberg invested $14.3 billion in the Start - Up Scale AI and put CEO Alexander Wang in charge of the lab. It also successively hired talent including Daniel Gross, co-founder of Safe Superintelligence, and Nat Friedman, former GitHub CEO.