Researchers from Oxford University have warned members of parliament that artificial intelligence (AI) could “kill everyone.” In an ordered inquiry by the Science and Technology Committee of the House of Commons, the Oxford researchers Michael Osborne and Michael Cohen, as well as Katherine Holden and Manish Patel from companies that deal with AI, spoke on its potential dangers and the import of its proper governance.
Osbourne has argued that over-reliance on AI is something that could carry humans into a new age of technological marvels and progress, leading to a phenomenon which he described as “bionic duckweed”. The professor noted that it leads to complacency, with people assuming the problems of the present will be solved in the all-too-murky future, and assuming a utopian version of AI that simply does not exist, as it is “meeting the goals we say, not the goals we want.”
Osborne was also quick to point out the various properties of AI that lend itself towards superhuman capabilities, namely in that of a capitalist economy that prioritises production. “AI can work 24/7, and it does not get distracted […] AI is scalable to a degree that humans are not.” The fundamental fear during the inquiry seemed to be what happens when an all-powerful artificial intelligence begins to decide for itself what it wants to do, calling forth images of Skynet from the Terminator series, or HAL 9000 from 2001: A Space Odyssey.
One member of parliament, Aaron Bell, voiced some skepticism at Osborne’s declaration of doomsday if AI was allowed to proliferate unchecked, asking “how realistic [did Osborne] think the bleak vision is?” Osborne replied decisively, comparing AI to nuclear weapons in power, and warning against potential arms races that could begin between countries trying to build the most AI, which Osborne noted was a “military technology” that could be used to control drones and kill combatants independently of any human intervention. “You do not just want to have a human dummy rubber-stamping decisions made by an AI…,” Osborne said.
But in voicing an actual timeline between the relatively faulty models of AI present today and the generative, transformative versions seemingly afforded by the future, both Cohen and Osborne were reluctant to give any firm timetables. Cohen related a story of Ernest Rutherford proclaiming that nuclear energy was impossible, only for it to be achieved less than 24 hours later. “It might look a lot like it does today months before [an AI paradigm shift]. Technological progress often comes in bursts.”