Really, the next Bond movie writes itself.
asserted
movie → write → 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.
asserted
version → destroy → threat
And what’s the new threat, except rogue AI?
asserted
threat → ’ → 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.
asserted
Lowden → ’ → footwear
And in any plausible scenario it would be the Americans plus Bond plus China against the prospect of AI going rogue.
asserted
AI → go → rogue
Jack Clark, an Anthropic executive and co-founder, agreed that leaving AI to be a “totally unregulated industry” was “rolling dice with immense risks”.
asserted
leaving → agree → risks
Very far from denying the charge, bosses of AI companies — including Elon Musk — agreed with the doom scenario.
asserted
bosses → deny → scenario
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.”
asserted
we → tell → this
They feel the magnitude.
asserted
They → feel → magnitude
Does that make us feel better?
asserted
us → make → ?
Altman himself, a nice-looking bloke who like every other tech oligarch often wears disarmingly informal T-shirts, shares our concerns.
asserted
who → wear → concerns
“It doesn’t take as much imagination as it used to for [us] to imagine how this could go wrong,” he said.
uncertain
he → take → imagination
“I think the world is right to be afraid of this.”
asserted
world → think → 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.
asserted
horsemen → call → risks
Anthropic chief executive Dario Amodei called for the pace of all AI development to slow and said governments should regulate the industry.
asserted
governments → call → industry
But that doesn’t square with the modus operandi of other tech companies.
asserted
that → square → 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.
asserted
versions → say → governments
So: the future of humanity should be left to him and the other bros.
asserted
future → leave → him
If it makes you feel better, Mark Zuckerberg of Meta says that market forces alone are enough to contain the threat of rogue AI.
asserted
forces → make → AI
Every AI lab has, he says, the ability and the incentive “to take its own actions” in making AI tools and models safer.
asserted
tools → have → actions
“Any lab that doesn’t focus on alignment will fall behind,” Zuckerberg said.
asserted
Zuckerberg → focus → alignment
“Labs face significant liability if their models cause harm, so they have a strong incentive to prevent this as well.”
asserted
they → face → this
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?
asserted
I → come → 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.
uncertain
it → seem → value
So, for fear of being sued or losing competitiveness, companies will try to avoid catastrophe.
asserted
companies → sue → 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.
asserted
liabilities → come → 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?
uncertain
it → sue → us
But would you trust them with the future of humanity?
asserted
you → trust → humanity
Some are suggesting we should.
asserted
we → suggest → ?
Both were speakers at this week’s Pro-Human Assembly in Washington and both demanded stringent regulations on AI.
asserted
both → demand → AI
Sanders called for co-operation with China — and as it happens,
Read More
asserted
it → call → China
Xi Jinping is visiting the US next week.
asserted
Jinping → visit → US
“A super-intelligent AI that escapes human control will not be an American problem,” Sanders said.
asserted
Sanders → escape → control
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.
asserted
He → compare → War
That is the only comparison with the present danger: the threat from weapons which could destroy whole cities.
uncertain
which → destroy → cities
In fact, we came pretty close a few times to that scenario.
asserted
we → come → scenario
But back then, the weapons were almost wholly under state control and there were rational men in charge of those states.
asserted
weapons → be → states
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.
asserted
AI → insist → good
Amodei, the Anthropic boss, helped formulate that encyclical.
asserted
Amodei → help → encyclical
Any chance he could pass a copy on to the other AI titans?
uncertain
he → pass → titans
The Aporia
