Britain Seeks to Build a Homegrown OpenAI Rival, Become a World Leader in AI (cnbc.com) 65
"The U.K is looking to build a homegrown challenger to OpenAI and drastically increase national computing infrastructure," reports CNBC, "as Prime Minister Keir Starmer's government sets its sights on becoming a global leader in artificial intelligence."
The government is primarily seeking to expand data center capacity across the U.K. to boost developers of powerful AI models which rely on high-performance computing equipment hosted in remote locations to train and run their systems. A target of increasing "sovereign," or public sector, compute capacity in the U.K. by twentyfold by 2030 has been set...
To further bolster Britain's computing infrastructure, the government also committed to setting up several AI "growth zones," where rules on planning permission will be relaxed in certain places to allow for the creation of new data centers. Meanwhile, an "AI Energy Council" formed of industry leaders from both energy and AI will be set up to explore the role of renewable and low-carbon sources of energy, like nuclear...
Britain plans to use the AI growth zones and a newly established National Data Library to connect public institutions — such as universities — to enhance the country's ability to create "sovereign" AI models which aren't reliant on Silicon Valley... Last month, the government announced a consultation on measures to regulate the use of copyrighted content to train AI models.
Britain plans to use the AI growth zones and a newly established National Data Library to connect public institutions — such as universities — to enhance the country's ability to create "sovereign" AI models which aren't reliant on Silicon Valley... Last month, the government announced a consultation on measures to regulate the use of copyrighted content to train AI models.
Good luck (Score:2)
Re: (Score:2)
Ah, but if cost is to high, then you get kw/h, or kilo-whines per hour. It's a very important metric to government officials.
Re: (Score:2)
The developers don't need to be co-located with the data centers.
The data centers could be located in Iceland (geothermal) or Quebec (hydro).
Deepmind is British. It's part of Google now and has 2500 employees in London.
Using bureaucrats to manage entrepreneurialism has a poor track record (Quaero, Solyndra, Japan's Fifth-generation project, etc.).
The UK should create a welcoming environment for all startups rather than "picking winners".
Re: (Score:2)
Afraid it's gonna take a dollar out of your pocket, bill?
A belief in the value of public sector computing is not "picking winners".
Re: (Score:2)
Afraid it's gonna take a dollar out of your pocket, bill?
The UK doesn't use dollars, and I don't live there, so it's not my money.
A belief in the value of public sector computing is not "picking winners".
Public subsidies for a "challenger to OpenAI" certainly is an attempt to pick a winner.
AI startups were a great investment ten years ago when OpenAI started.
They are not a good investment today because you'll be ten years behind.
Ten years from now, we'll know what was a good investment now. But we can only guess at what it will be. Bureaucrats are terrible guessers.
You can't predict the future by looking in the rearview mirror.
Re: (Score:2)
AI right now is a fad. It would be like the government investing heavily in the dotcom sector in the late 90s. Sure, a couple big winners, but then the taxpayers lose out on the collapse of pets.com.
You know how flimsy AI prospects are given the absolute glut of marketing for it. A decent product doesn't need hype.
Re: (Score:2)
AI right now is a fad.
That's the problem when governments play VC.
They invest in what's hot right now, not what will be hot in ten years.
Re: (Score:2)
So... just fund basic research, old school style. The drawback is that the billionaires don't get direct payments, which doesn't bother me much.
You want smarter citizens, then invest in citizens. You want dumb labor, then invest in billionaires.
Re: (Score:2)
...yeah, far too much. We're far too reliant on gas to get the price down to something sensible. We also have a lot of taxes on the price for environment reasons, which you could argue are bad, but are far less so than the over-reliance on gas.
Suffice to say, the planning laws are not what's preventing datacentres being built. It's actually not too hard to build one if you really want to. It's just that it'll be hard to make it economically viable, given that the people using it don't need to be anywhere ne
Re: Good luck (Score:2)
Read 'homegroom' (Score:2, Funny)
must be all the other news about Brittain.
Re: Read 'homegroom' (Score:1)
Re: (Score:2)
Yeah why would we want to protect children? If Musk is doing that then we should be anti-children!
If Musk were trying to 'protect the children' in question he'd probably do better by encouraging adoption of the recommendations of the previous enquiries into the scandals, rather than calling for a(nother) national enquiry. Calling out the previous government's - and its many 'leaders' over the last 14 years, albeit still only the one party, - incompetence and lack of action, rather than hating on the current one, one that's been in power for less than a year, would also strike me as a more appropriately
Re: (Score:2)
Musk is anti-children, look how he treats his own. "Pedo" is weapon to use against enemies, not to protect children.
Re: (Score:2)
Fortunately for us, Musk is so clearly, blantantly anti-children, that he is even anti-his-own-children. He turned on his own transgendered daughter, to the point of saying about her "the woke mind virus killed my son". Publicly, and completely unnecessarily, in comment of a comment.
It's one thing to not understand your child, but it takes an asshole to not support their own child, and I run out of reasonable names when we get to what a person who won't even respect a decades old decision by your child. In
Re: (Score:2)
Musk is a Johnny-come-lately who knows nothing about the issue but still want to stick his.. finger into the mix. It's a ten year old story. Many inquiries have been done. He's following the Trump model, where he reads some headlines then decides to post about it before reading the stories or getting the details (he'd fit right in to Slashdot).
If Musk cares about the children, he should demand the release of more Epstein documents.
This is getting weird (Score:5, Interesting)
Somebody is advising governments that being the first to develop a really useful AI is worth the government's attention and taxpayer's money. Building and training these models is going to be chewing through significant amounts of energy.
Yet despite all this, we still don't have a theory of mind to tell us why the AI we can create today doesn't even have the intelligence of a small insect brain. And if we did, and could make AI as intelligent (or moreso) than we are, we have plenty of people who have rational reasons for thinking that's a bad idea.
So why are we all rushing headlong towards a goal we don't yet know how to attain, at significant expense, and with massive potential downsides (like destroying our economies overnight once AI can replace everybody and we don't have a backup economic system available)?
Re:This is getting weird (Score:4, Insightful)
It's free taxpayer money for influential people who will pay it back with donations to Labour.
The UK is a joke. It periodically announces that it is going to become world leader in something, and then doesn't do even a tiny fraction of what is necessary to make that happen.
We have the ambition of China, but no ability.
Re: (Score:2)
It's free taxpayer money for influential people who will pay it back with donations to Labour.
The UK is a joke. It periodically announces that it is going to become world leader in something, and then doesn't do even a tiny fraction of what is necessary to make that happen.
We have the ambition of China, but no ability.
Sadly this, it's not the first time as Rishi Sunak tried this a year or so back. Whenever the govt wants to distract the media.
Can't we just stick to what we're good at which is banking, complaining and closet racism?
Re: (Score:2)
We need more than that, at least 95% of us do. I don't know what the solution is, if there even is one.
Re: This is getting weird (Score:2)
Re: (Score:2)
So why are we all rushing headlong towards a goal we don't yet know how to attain, at significant expense, and with massive potential downsides (like destroying our economies overnight once AI can replace everybody and we don't have a backup economic system available)?
Because all the important nations have an AI. If the UK doesn't get an AI, it'll stop getting invited to the best parties.
Re: This is getting weird (Score:2)
Re: (Score:2)
Yet despite all this, we still don't have a theory of mind to tell us why the AI we can create today doesn't even have the intelligence of a small insect brain.
Sure we do. The LLMs are fancy autocompletes, and they behave exactly as expected -- for example, they can write some near-perfect citations in support of the point they're trying to make, but why would you expect them not to be fictional? Even so they're great for various purposes, as long as you understand how they do and don't work.
Re: (Score:2)
I think you're giving them much too little credit, since (even if your characterization is correct) autocomplete is finding a path through a conceptual or linguistic solution space that meets certain demands. How good the system is is determined by the demands it can meet. Modern chatbots can imitate individuals. They can explain chemical engineering, at least like a first year student. They can write Rust. Most adults can't navigate their way through these solution spaces, even if they have time to train.
Re: (Score:2)
The less you know, the more impressed you are, and you know little.
No one doubts that computers can do these things theoretically, but they cannot do them except at the most cosmetic level now. It's profitable to pretend they can, though, and it's easy to fool people like you. The goal is to get money by lying, and this is just a political knee-jerk reaction to the lies. When you lie enough, politicians are among the rubes that fall for it.
"You call it autocomplete, I call it fuzzy data generation."
But y
Re: (Score:2)
Wow. Okay. You're wrong, but I'm not going to explain because you're kind of a jerk. Like, the kind that was taught he's not supposed to be a jerk, so he steps up as close as he thinks he can get to the line.
Re: (Score:2)
I'm sure most humans could use Google to look up snippets of code as well as an LLM. The magic of LLMs is that they have a sample of nearly every bit of published human knowledge, with enough pattern matching to make new things fit in. It gives the illusion of knowledge. Easy Problems That LLMs Get Wrong [arxiv.org]
Though I do foresee a combination of LLM plus a model-based algorithm having much greater potential.
Re: (Score:2)
Thank you, that was a lot of fun to read. I've seen chatbots get basic things wrong that betray the fact that they don't understand what they are saying, but my examples aren't as fun. (And I'm using "understand" to refer to the ability to use information.) And I also feel it would be a shame if we don't go beyond LLMs, considering the extreme measures companies go to to train and run them in training and running them while getting such inconsistent results. It seems like this must be the wrong approach, un
Re: (Score:2)
Somebody is advising governments that being the first to develop a really useful AI is worth the government's attention and taxpayer's money. Building and training these models is going to be chewing through significant amounts of energy.
Yet despite all this, we still don't have a theory of mind to tell us why the AI we can create today doesn't even have the intelligence of a small insect brain. And if we did, and could make AI as intelligent (or moreso) than we are, we have plenty of people who have rational reasons for thinking that's a bad idea.
So why are we all rushing headlong towards a goal we don't yet know how to attain, at significant expense, and with massive potential downsides (like destroying our economies overnight once AI can replace everybody and we don't have a backup economic system available)?
The owner class is obsessed with AI because they see it as the ultimate avenue to complete and forever control over all of humanity. When the owner class becomes obsessed with something, they convince the governments they own to throw resources at whatever it is they are obsessing over. While we may hope that something changes to shake them off this path, I think we're going to be stuck with LLM driven data aggregation and regurgitation routines as AI, and most likely as enforced "wear this device or leave
With what tech base? (Score:5, Interesting)
When I pull up a list of international tech companies by market cap, the only one in the top fifty that's UK-based is ARM. Granted, market-cap is not the only indicator of performance, but it's not exactly helping when there aren't a lot of massive companies with significant R&D labs with large budgets that can contribute to the effort.
A pretty common trajectory for tech companies is to grow to the point that they feel the need to consolidate their various R&D projects to specific college-like campuses, to get several years or even decades of research contributing to new products, then for venture capitalists to get in and complain about the costs to run these incubators on account of the short-term costs while ignoring that nearly all of the company's products from the last decade came from those efforts, leading to closures.
Right now the UK doesn't seem to have large firms in this R&D-park model. The UK has colleges that are strong in their respective computing disciplines, but without the money from companies with long-term R&D in mind, I just don't see them managing to develop even to the standard of what's coming out of other parts of the world, let alone to exceed it, and with Brexit it may be difficult to attract foreign talent as well.
Re: (Score:3)
I don't have mod points but that is an incredible statistic.
The resources of private industry to develop AI dwarf even that of the US government (I don't mean procuring GPUs, I mean luring the best and brightest by paying thousands of people $500K+ each), which in turn dwarf that of the UK government. That's before we even start talking about regulation such as GDPR. I hate to say it
Re: (Score:2)
"I mean luring the best and brightest by paying thousands of people $500K+ each..."
Ah, the Silicon Valley perspective. You're already disqualified from the conversation.
Re: (Score:2)
Re: (Score:3)
Headquartered in UK, A Japanese investment company owns it and it's traded on the Nasdaq.
Not much of it is UK based.
Re: (Score:3)
The UK is a low wage economy with Brexit cutting off access to a lot of talent and cooperation. Being outside the EU means that tech developed here will need to be exported - the domestic market isn't big enough to justify that much investment in R&D. That means recertification and maybe tariffs, to a market that favours domestic stuff.
The only reason that Britain can boast about "investment" is property and selling off our assets.
This isn't going happen and everyone knows it. It's a joke to even announ
Re: With what tech base? (Score:2)
If the UK is a low wage economy then I wonder what Pakistan is. Or Nigeria.
Re: (Score:3)
Low wage compared to Europe and the cost of living.
Re: (Score:2)
As with all statistics .. there is far more to the picture than you suggest.
Google's AI research is done in London, for example. Just because the companies are listed on the US stock exchange and their headquarters are in the US, doesn't in any way show where the actual work is happening.
The UK suffers from a cultural snobbery towards technology that tends to direct UK capital away from such investments, it does not suffer from a shortage of high quality research (both commercial and private), funded from a
Re: (Score:2)
You're assuming that R&D and tech developments can only come from the top 50 firms based on market cap. That is not only a shitty assumption, it actually couldn't be more wrong. The vast majority of tech development and R&D isn't driven by large corporations. It's from minor companies and startups which at some point later get eaten by massive corporations.
Oh maybe you need finance? Well that's often easier to get outside of a major company for a pet project. Investors are typically easier to convin
Re: (Score:2)
Except ARM is mostly not British now, it's owned by Softbank as majority stakeholder. Whoops.
The snag is that UK is relatively very small, and US is huge. That means despite UK being the primary science and tech power house before WWII, the US dominated. Computers were very much a British thing, but the US dumped huge amounts into it because of the cold war, and it had more money to dump than UK did.
And today, split off from EU, the UK is even smaller in influence. Whoops.
Mixed signals (Score:2)
So Britain wants to encourage domestic AI while removing the fair-use copyright exemption.
What are they thinking? Why would any company train their AI models in Britain, when they can do it in places with no restrictions on training data (America, Dubai, Singapore, Japan, ...)
Re: Mixed signals (Score:1)
Re: (Score:2)
You do not think regulation of intellectual property is a function of government? And you think "fair-use copyright exemption" is world-wide?
"What are they thinking?"
They are thinking that it is the role of government to do government's role.
"Why would any company train their AI models in Britain, when they can do it in places with no restrictions on training data (America, Dubai, Singapore, Japan, ...)"
Because where you train an AI model doesn't determine if you are violating intellectual property laws?
Remote services that you're trying to localize? (Score:2)
Why do you care where the data centers are? The results shouldn't differ THAT much, with good network design and capacity.
This should be an economics based decision, not a political one. Are they really that stupid, and believe that doing IT in data centers is going to benefit them or their citizens?
How can the government afford it? (Score:1)
The UK government can't keep its pensioners warm. How can it afford the voracious appetite of an LLM?
Re: (Score:3)
Re: How can the government afford it? (Score:2)
The UK government has chosen to stop handing subsidies to pensioners that can afford heating. FTFY
Ahahahahahaha (Score:2)
Some former empire thinks thy can just _decide_ to become a market leader in something. That is not how it works. That is how abject stupidity and lack of education works, though.
Deepmind was in London (Score:2)
The first step is hard (Score:2)
Building a time machine to go back 80 years to alter laws and culture to establish cradle innovation in some part of post-WW2 Britain.
Before or after... (Score:3)
I'm still waiting for the UK to become Global Britain and a world leader in ... I honestly can't even remember what it was they were holding up as a Brexit benefit. If the UK do become a global leader in AI, it'll be the result of a startup's innovation, or university research. It won't be because the King's incompetent government wants it.
Re:Before or after... (Score:4, Interesting)
I'm still waiting for the UK to become Global Britain and a world leader in ... I honestly can't even remember what it was they were holding up as a Brexit benefit. If the UK do become a global leader in AI, it'll be the result of a startup's innovation, or university research. It won't be because the King's incompetent government wants it.
Used to be banking, medical and pharmaceuticals and heavy machinery (seriously, UK produces a lot of heavy plant, bulldozers, diggers, jet engines, turbines, that can only really be built to a high standard in the developed world) but Brexit has been killing it all as "quelle surprise" our competitors are similar developed nations most of whom are across the channel, the very market we stupidly cut ourselves off from.
Thanks to our friends across the pond, we're not even world leaders in dreadful racist politicians any more. Yet another failure from Farage.
Re: (Score:2)
Re: (Score:2)
Apparently sequoia trees?
Re: (Score:2)
It means industrial equipment, quite often fixed. Reactors, pumps, hydraulic systems, Separators, sieves, material conveyors, mining equipment.
I have a c3po AI from 2002 (Score:2)
And I want a pony (Score:1)
DeepMind? (Score:2)
Isn't that what DeepMind is?
Or are they looking for an investment scam, garbage producer, because that is what Open AI is,
GPT-4 is garbage, but it attracts a lot of investment.