Artificial intelligence poses a “Hiroshima”-style risk to humanity if governments do not agree to curb how it is developed, the foreign secretary has warned.
Yvette Cooper urged countries, including the US and China, to agree international rules for AI, telling the Guardian she believes the issue will dominate foreign policy over the next two years.
And even if governments decided they’d like to try to make it happen it still wouldn’t happen.
Now, OK, this is me being flippant, but what you need to make an AI is coders and server time. It’s simply not possible to control those two in any useful sense. Therefore AI cannot be controlled. QED.
Whatever politics has to say about it.
“The whole aim of practical politics is to keep the populace alarmed (and hence clamorous to be led to safety) by menacing it with an endless series of hobgoblins, all of them imaginary”
H.L. Mencken
Yes, but that’s only half the story. The other aim of practical politics is to buy votes so that you can spite those you dislike, with their money.
Like rubbing our noses in diversity Norm?
You could say that, Addolff.
They chose a particularly useless specimen to unleash that particular hobgoblin. WTF does Yvette Cooper know about AI and its pluses & minuses?
I think Yvette Cooper is the hobgoblin.
“Artificial intelligence poses a “Hiroshima”-style risk to humanity”
How? it’s just code giving you answers. Now, it’s probably worth not plugging AI into airlines, but we already have regulations around that.
The whole thing is massively overhyped.
Anyone else remember the grey goo from the nanobots scare stories?
And the Millennium Meltdown, or whatever it was, that was going to lead to the complete collapse of the financial sector, causing food riots and a zombie apocalypse.
You forgot to mention the planes falling out of the sky.
One a personal note, I was tasked with going into work on New Year’s Day 2000 to check that all the company’s PCs were still OK. I did get my travelling costs refunded.
I was ordered to stay on the job til after midnight to make sure all was running right.
I ran an accounting system. NO ONE was doing accounting at 0:00:01 on 1 January 2000. Not for another 56 hours.
ISTR that King Chuck was very fond of that nonsense.
WB, you know there’s code in airliners. And most people don’t know where the line is between code and AI anyway. However, we have has a Hiroshima event and life didn’t even completely stop in Hiroshima. Addolff is right, it’s a Menckenian Hobgoblin. And anyway, it’s the hun you don’t see that gets you
The people complaining are the people doing reports and spreadsheets. Which are often bullsh*t jobs. The complaint is that the jobs are being replaced by bullsh*t of a different origin.
I have said before. The issue isn’t really AI. It is ML robotics that will cause a technological holocaust. AI doesn’t exist and may take a century to get a foothold.
Having robots, with the ML that we call AI, will wipe out the workforce. Unskilled people are easily replaced and once the balance changes we all fall off. Then what do the skilled people need to do if there is no productivity improvements necessary or any new innovation. So most of them will go as well and so it goes up the ladder.
AI/ML/robotics aren’t free, certainly not while we’re having stupid arguments over what energy production we’re ‘allowed’ to have, so it’ll replace people who add the least to any process. How big a slice that is largely depends on how much growth can get past govt policy.
However, any process that requires human oversight will still bottleneck. I’m reminded of being on jury duty for a fraud case, and watching the barrister laboriously read (ie speak) large chunks of account ledgers into the record. Same with my day-job, verifying the lumps of code AI spits out is faster than writing it myself, but not catastrophically so. Esp since there’s only so much focussed thinking you can exercise in a day. And there are usually a few optimisations, conventions, and user expectations to herd the AI into implementing.
Toy apps, funky websites and chatbots can afford to be 80%, even 60% accurate.
My view is that the workforce will get smeared out over a much larger production ‘estate’, in countries that get on board with that. Same as industrialized farming. Other countries will struggle to be competitive.
Possibly we’ll see more client-facing jobs paid for by the savings. That’s a thing people value, so if the AI optimises everything else it’ll become an affordable marketing advantage.
Not gonna be painless, and it’s possible I’m being way too optimistic, and most countries just can’t grow to soak up the excess labour. That’d be very bad.
And doesn’t cover the original point, which is that AI is easy enough to adopt that govts will struggle to regulate it. Hell, we’ve seen that govts have incentives *for* unregulated AI, in the Trump/Anthropic farrago. They will be built, they will leak, and even the growth in regulated stuff will beat paths that benefit people trying to work around those regs.
“watching the barrister laboriously read (ie speak) large chunks of account ledgers into the record.”
Either spoken testimony is treated differently than written exhibits or the barrister is protecting his hourly fee.
Possibly it has more impact if the jury hears it. Though I doubt that; mostly it would irritate many.
The average jury* has neither the competence nor inclination to understand ledger entries, even if they were given the time to understand them properly as written evidence. If the barrister is merely reading out entries, he’s also wasting everybody’s time. A competent prosecuting counsel will explain the meaning and significance of the figures to make his case.
I’ve always assumed the purpose is that verbalising the information in a room with a judge, jury and opposing barrister is the least-worst way of ensuring everyone has heard the same information, and nothing more or less.
You could refer to a block of transactions and just let everyone read at their own speed, but you have no idea what they actually do. Whether they read more, or less, whether they misread something, or look at the wrong block entirely. Whereas what’s been said is what has been said, and very likely heard by everyone. And correctable by the judge or other council.
Won’t rule out job protection, though, or even just sclerotic, desperately inefficient processes without the technical heads around to build the smarter ones. And if anyone knows the specific reason, I’d love to know that.
Years ago I was involved in a huge corporate deal that included a French company as one of the parties. French rules were (dunno if they still are) that a French lawyer had to read the contract out loud to the client, so that they knew what was in it and could ask questions if they weren’t sure about anything.
This may have been fine for an ordinary contract, but not for one of over a thousand pages.
There were French lawyers who were trained in speed-reading, so we hired one of those; she did the whole thing in less than an hour.
Complete waste of time of course, but it does show that people can usually find a way to get round these rules, although at the cost of wasted time, effort, training and money.
We’ve generally replaced the unskilled people with (some) of the foreigners we’ve imported. I suspect that replacing those with machines won’t be all that hard.
Except for the whining from certain people who want those foreigners around for reasons they don’t want to say out loud.
At last someone talks some sense. And a machine will be writing the code for those robotics.
Find it fascinating how many of you can’t see past the parlour trick that is Large Language Model. But what can you expect from people who are basically desk jockies.?
Who is deploying this anywhere? Answer: no-one.
There’s a lot of talk about AI robotics. I have not heard of a single deployment of such a thing.
I would say that unskilled people are actually harder to replace than the extremely skilled.
Since most unskilled labor is manual you need a physical machine in there to do the work. That’s expensive. Right now the solution is to have a teleoperated robot doing the work and the worker in Bangladesh in the hopes that some time in the future ML can run the robot for you.
This has been the secret scam behind more than one recent ‘home robot’ startup. Like Micro1 and Shift.
They’re being helped along though by the ever-increasing minimum wages making unskilled labor more and more expensive.
The problem with the current generation of AI is that it isn’t really coded so much as brewed from vast amounts of information and then the way it outputs things is coded out of it with a lot of tweaking to keep responses “within parameters”. That gives a reasonable facsimile of a “human” response, but it’s not really even that, little more than a very expensive ventriloquists dummy. Someone once described the current version of AI as “a stochastic parrot”, that’s pretty much spot on. As Lord T says, I’m more concerned about the job devastation to come from Machine Learning robotics. Especially if the price point for a unit is $30,000 as
OzElon the Great and Powerful has suggested.The US AI companies have talked themselves into federal regulation, which I think they’ll come to regret.
I know a sales pitch when I see one, such as AI companies talking up the “dangers” of AI. “You see, our products are so amazing they need (gasp) LAWS” is honey in the ears of investors betting a graphics card manufacturer is worth trillions.
Not sure they bother making graphics cards anymore.
The goal of these companies is building themselves up to a point that they can be sold off to a load of idiot retail investors. Everything is about marketing the companies as massive potential investments to people who don’t understand the technology but fear missing out on huge juicy gains.
Six months after the companies are sold, and all the pre-IPO investors can start selling off, they don’t give a toss what happens to them.
what you need to make an AI is coders and server time. It’s simply not possible to control those two in any useful sense.
This is the segue into them trying.
There are plenty of roll your own LLM AIs out there that are free now. You can get some amazing results with hardware costing several thousand pounds. (eg M5 Pro Macbooks with 48GB of storage and 2TB ssd).
It can also be trained by yourself rather than the silicon valley leftie types, so its biases won’t reflect the views of the PTB.
Very dangerous stuff indeed (to the elites that is).
An empty cab drew up and Yvette Cooper got out.
Can anyone think of anything Pixie has done?
Annoy me.
“. .she believes the issue will dominate foreign policy over the next two years.”
And I believe her. Because the same useless c###s occupying diplomatic positions and utterly unable to answer the man-made questions will deflect by talking about AI so that nobody notices how useless they are.
I find that a tasteless thing to say.
Oh, yeah. Governments can do that. Based on decades of observation, I’d say the more government gets involved, the SOONER the bomb will drop.
Dominate? Rubio will tell her to FO. But Britain can have fun playing AI diplomacy if it wants. It’s been having fun BY ITSELF with its Net Zero game. Why not AI, too?
AND ?!?! WTF ?!?!
Sure, China are absolutely going to sign up to international limits to AI, and not have their fingers crossed behind their backs the whole time….
“That’s a frightfully complicated computer program with millions of lines of code you’ve built, now I’m going to tell you how it can be FAIR and then solve the middle east.”
Yeah, no.
A member of the same government that wants Britain to be a world-leader in AI development calls for there to be limitations on AI development. It’s the same sort non-joined-up thinking that thinks it’s possible for Britain to be a world-leader in AI while also having the highest electricity costs in Europe owing to the same government’s environmental policies.
They’ll endeavor to stymie it through opposition to building data centers.
Which all have their own gas power plants anyway. CO2? Wossat?
In an essay covering her thoughts on everything from emerging technology to Palestine, Cooper said the world was at a dangerous moment, not least because of what she sees as the permanent withdrawal of the US from its role as a global arbiter.
Lol. Mrs Balls has never expressed an original thought in her life, so who’s playing along with the idea this silly cow is a thinker?
Chatham House, obvs, but everyone else has noticed the lanyard classes are malevolently incompetent fuckwits. Labour’s cool new idea to censor the internet involves forcing YouTube to shovel BBC content at you against your will. I don’t think that’s going to work out the way they think it will.
And have you pay the BBC licence for it, too.
Dunno, the British government managed to get Apple to remove its advanced encryption between UK iPhone users’ and Apple’s servers (so Apple could avoid giving the UK govt a backdoor into the phones). And mysteriously, the UK became the first country that Apple forced its ID check / age verification on. It seems the UK govt has a lot of influence over US tech companies. Where is Trump?
YouTube is currently behaving like it expects the propaganda / censorship regime to be enforced.
I’m thinking of starting a collection of Very Strange Conspiracy Theories.
One I saw recently was that there was no A-bomb at Hiroshima, there was a standard US fire-raising raid plus one bomber that released radioactive trash from the Manhattan Project.
So we are asked to believe that this masquerade fooled the USSR even though the Project, and the Federal government more generally, were stuffed to the gills with commie spies.
a collection of Very Strange Conspiracy Theories
dearieme, please don’t use commie language. They are Very Strange Theories.
How did they fake Ivy Mike and Castle Bravo?
“don’t use commie language”: it’s CIA language. I can’t swear that that’s a different thing.
What are Ivy Mike and Castle Bravo?
Very big bangs indeed. Castle Bravo was 3x more powerful than predicted. It was the USA’s first solid-fuel hydrogen bomb.
Isaac Asimov, clever bloke, already thought about the problem of regulating artificial intelligence in the 50’s. But Yvette Balls is the cleversilly woman from RoboCop 2, who would like to fill the robot’s electronic brain with a load of social justice crap until his head explodes
Not the full list and honestly how does one enforce it?
Yeah. It’s like . . . until it does.
Except the main point of most of his stories was how the laws malfunctioned either preventing a robot doing something desireable or finding a loophole and doing something bad.
The fundamental problem which can never be overcome is that what I want from universal rules is likely to be what benefits me and therefore not what others want. Even when what I want does not benefit me, it so easily falls to the tyranny of good intentions.
Absolutely, like the zeroth law
Except the main point of most of his stories was how the laws malfunctioned either preventing a robot doing something desireable or finding a loophole and doing something bad.
Exactly! Charles read and understood the source material. The rest of you see me after class.
Asimov – again, clever bloke, I think underappreciated because he wrote pulp sci fi for dorks – not only foresaw the need to regulate autonomous machines. He drafted, in the highest level programming language there is, English, the essential code/regulations any reasonable non-sociopathic humans would expect of their potential killbots. And he worked out how humans, robots and even semantics itself might defeat the best intentions of restraining the capabilities of new technology.
And his books sold millions of copies. Nobody will read Yvette Cooper’s half baked ignorant platitudes about AI, except a handful of journalists, political researchers from the other parties, and LLMs. Asimov was a techno-pessimist, even his perfectly planned societies ruled by autistic atheist geniuses in Foundation couldn’t withstand the ravishes of history and humanity. Even his shiniest and cleverest robots might kill you.
But not this portable computer
What does ‘Hiroshima style risk’ mean?
Because Hiroshima wasn’t close to a risk to humanity.