Microsoft is the latest company to announce efforts to develop superintelligence, a hypothetical form of artificial intelligence that would surpass human capabilities. However, some experts fear that such a technology would be impossible to control and, therefore, could endanger humanity’s future.
Microsoft is attempting to ease those concerns with its MAI Superintelligence Team, with the stated goal of creating a “humanist superintelligence” designed to “remain controllable, aligned, and firmly in service to humanity.”
“Our mission is to build the safest, most capable systems that keep humans in control and deliver real-world impact,” the company said last week.
Download the SAN app today to stay up-to-date with Unbiased. Straight Facts™.
Point phone camera here
Microsoft’s AI chief, Mustafa Suleyman, says the company’s aim is not to chase an “infinitely capable” AI, in contrast to others pursuing superintelligence, such as Meta. Instead, Suleyman says, Microsoft will focus its superintelligence efforts, at least to begin with, on medical diagnostics. This focus, theoretically, would allow Microsoft to retain control.
“We’ll have expert-level performance at the full range of diagnostics, alongside highly capable planning and prediction in operational clinical settings,” Suleyman said.
The MAI Superintelligence team also plans to expand into areas such as education and renewable energy.
‘We need to have a clear conversation’
Reactions from AI experts are mixed. Vincent Conitzer, a professor of computer science at Carnegie Mellon University, told Straight Arrow News that he believes Suleyman’s concerns were genuine. But given that superintelligence, by design, exceeds humans’ capabilities, Conitzer questioned Microsoft’s use of the term.
“It seems that what they want to do, or at least Suleyman wants, is precisely not to create superintelligence, at least superintelligence in the sense of an intelligence that exceeds our own abilities across the board,” Conitzer said. “The risk is that by calling this goal ‘superintelligence,’ which may be good for Microsoft’s marketing, we all become even more confused about what it is that we’re actually talking about. We need to have a clear conversation about this as a society, instead of having this be driven by how companies want to market their AI initiatives.”
While debate continues over whether superintelligence is even achievable and controllable, others question Microsoft’s plans as well.
Max Tegmark, a physics professor at the Massachusetts Institute of Technology, suggests that Microsoft should support strict AI regulation to demonstrate its genuine understanding of the potential dangers of superintelligence.
“If Microsoft is serious about not racing to superintelligence before the control problem has been solved, they should come out in support of a law requiring the same from their competitors, rather than opposing all binding safety standards,” Tegmark told SAN.
Conitzer also feels that Microsoft’s messaging regarding superintelligence appears contradictory.
“Corporate communications are too likely to try to have the cake and eat it too, on one hand arguing that the AI will be as powerful as imaginable from the perspective of economic or other value, but on the other hand dismissing the risks and saying, ‘Oh that’s not the kind of AI we’re going to be building’ when the conversation turns to risks,” Conitzer said. “That’s not a good basis for the conversations that humanity needs to have about this. Let’s be clear about what we’re talking about and what we’re actually building.”
Suleyman did not respond to a request for comment from SAN.
‘Broad scientific consensus’ sought
Conitzer and Tegmark are among several hundred AI experts, policy makers and celebrities who signed an open letter last month calling for a ban on superintelligence development.
The ban, according to the letter, would not be lifted unless a “broad scientific consensus” is reached that superintelligence will be developed “safely and controllably” and with “strong public buy-in.”
Polling published with the letter showed that 64% of Americans agree, while only 5% feel that superintelligence should be developed as quickly as possible.
