Really, the next Bond movie writes itself. Like From Russia With Love, in which 007 is on the same side as the Russians against Spectre (a megalomaniac outfit with no national affiliations ready to destroy the world), the new version would unite old enemies against a common threat. And what’s the new threat, except rogue AI? Alas, there’s no bringing back Sean Connery for Bond (nope, Jack Lowden doesn’t cut it) nor the fabulous Lotte Lenya as Rosa Klebb, she of the toxic footwear. And in any plausible scenario it would be the Americans plus Bond plus China against the prospect of AI going rogue.
That prospect has got scarier in the last week, what with a former Anthropic employee saying AI could destroy humanity within 10 years.
Jack Clark, an Anthropic executive and co-founder, agreed that leaving AI to be a “totally unregulated industry” was “rolling dice with immense risks”. Very far from denying the charge, bosses of AI companies — including Elon Musk — agreed with the doom scenario.
Mustn’t worry, though. Sam Altman, head of Open AI, told a conference in San Francisco this week that: “The world should trust that we are going to do the right thing because it’s the right thing and we feel the magnitude of this.” Right. They feel the magnitude. Does that make us feel better?
Altman himself, a nice-looking bloke who like every other tech oligarch often wears disarmingly informal T-shirts, shares our concerns. “It doesn’t take as much imagination as it used to for [us] to imagine how this could go wrong,” he said. “I think the world is right to be afraid of this.”
The regulation debate
The obvious solution — if you can call a solution bolting the stable door as the four horsemen of the apocalypse disappear over the hill — is state regulation of tech firms and co-operation between the US and China to contain the risks. Anthropic chief executive Dario Amodei called for the pace of all AI development to slow and said governments should regulate the industry.
But that doesn’t square with the modus operandi of other tech companies. Nvidia founder Jensen Huang, a more combative type than Altman, said at the same conference that AI firms should decide all by themselves if new versions of the technology should be released rather than being managed by outside forces like, say, democratically elected governments. So: the future of humanity should be left to him and the other bros.
If it makes you feel better, Mark Zuckerberg of Meta says that market forces alone are enough to contain the threat of rogue AI. Every AI lab has, he says, the ability and the incentive “to take its own actions” in making AI tools and models safer. “Any lab that doesn’t focus on alignment will fall behind,” Zuckerberg said. “Labs face significant liability if their models cause harm, so they have a strong incentive to prevent this as well.”
If the worst comes to the worst, the financial liabilities of a tech company will be the least of our problems
Can I just dwell on that one? Zuckerberg, genius as he is, seems to be saying that if there were, say, a catastrophe with rogue AI, the kind that the Anthropic whistleblower said could destroy humanity, then the company concerned would be liable for damages; it would also fall behind its competitors in terms of stock value. So, for fear of being sued or losing competitiveness, companies will try to avoid catastrophe. The obvious riposte is that if the worst comes to the worst, the financial liabilities of a tech company will be the least of our problems. As a regulatory mechanism, the possibility of companies being sued for safety breaches works well for, say, household appliances; may I gently suggest it is inadequate to prevent self-improving technology from destroying its creators, along with the rest of us?
Sanders and Bannon united
Look at the top half dozen or so tech bosses: Zuckerberg, Altman, Musk and the rest. They are bright, driven, competitive, entrepreneurial. But would you trust them with the future of humanity? Some are suggesting we should.
There are few issues that unite Left-wing senator Bernie Sanders and former Trump strategist Steve Bannon, but on this they agree; regulation of the tech companies by government is the way to go.
Both were speakers at this week’s Pro-Human Assembly in Washington and both demanded stringent regulations on AI. Sanders called for co-operation with China — and as it happens,
Xi Jinping is visiting the US next week. “A super-intelligent AI that escapes human control will not be an American problem,” Sanders said. “It will not be a Chinese problem. It will be humanity’s problem.” Quite so.
He compared a putative Trump-Xi agreement with the efforts made by Ronald Reagan and Mikhail Gorbachev to avoid nuclear conflict at the height of the Cold War. That is the only comparison with the present danger: the threat from weapons which could destroy whole cities. In fact, we came pretty close a few times to that scenario. But back then, the weapons were almost wholly under state control and there were rational men in charge of those states. Compare and contrast.
The one person who has been consistently sane on this issue is Pope Leo, whose encyclical, Magnifica Humanitas, insists that technology is never neutral and that AI must be “disarmed” to become a servant of the common good. Amodei, the Anthropic boss, helped formulate that encyclical. Any chance he could pass a copy on to the other AI titans?
Melanie McDonagh is a columnist at The London Standard
