UK's new thinking on AI: Unless it's causing serious bother, you can crack on

The UK government on Friday said its AI Safety Institute will henceforth be known as its AI Security Institute, a rebranding that attests to a change in regulatory ambition from ensuring AI models get made with wholesome content – to one that primarily punishes AI-abetted crime. "This new name will reflect its focus on serious …

[Wiretrip]

Four simple rules

Said it before and I'll say it again - 3 things need to happen:

1) Make the AI companies liable for the output of their models; treat them like publishers.

2) Make them pay for their training material,like students do; remove their 'fair use' exemptions.

3) Make it *illegal* to give a model unsupervised or direct control of its host environment or another system.

4) Make the users of any AI system legally responsible for any decisions it makes concerning a person, e.g. Prison tariff, hiring decision etc. Ensure that a human has a veto and oversight.

[Anonymous Coward]

Re: Four simple rules

Yeah, nah.

1) Won't work...because anything is a weapon in the wrong hands. If you started to try and blame a DIY store for all the bad stuff that might happen if they sell someone a hammer, all that will happen is that DIY stores won't sell hammers, or worse they will sell "safety hammers" made out of cotton and fluff which can't hammer a nail...which punishes the majority of people that use hammers correctly. The penalty for AI misuse should land squarely on the user not the AI.

2) All knowledge should be free...even to students. Personal information, yeah they should pay for that in the form of a tax.

3) That removes a lot of the use cases for AI. I think what you're getting at here is that AI should not be able to make critical decisions. Somewhat agree, but it depends on the decisions the AI is making and what it is they control...nuclear silos...hell no...ordering milk when I run out in the fridge, why not? There are likely decisions that humans are unable to make for ideological reasons or refuse to make for malicious reasons that an AI can make without any regard for ideology and without malice that ccould be generally beneficial, until we test it, we don't actually know though...decision making based on facts without distortion is potentially one of the key benefits of AI...we need to establish whether AI being in control of things is actually bad before we make any assumptions. After all, just because someone disagrees with a decision, it doesn't necessarily follow that the decision was bad...we don't have any concrete examples of an AI making a decision that was inherently bad...yet...assuming that an AI will simply make bad decisions all the time is just tinfoil hate risk aversion at this point...because we just don't know...ironically, fearing the unknown has led to some pretty awful results in the past and ironically, some pretty poor human decisions.

4) Somewhat agree. Checks and balances make sense, we already do this when other humans are involved in important decision making processes. However, checks and balances normally involve many humans and several experts...giving one person a veto is probably worse than unfettered AI.

[Pascal Monett]

Knowledge should be free, I agree.

Data is not knowledge.

Don't confuse the two.

[Anonymous Coward]

I didn't...

"All knowledge should be free...even to students. Personal information, yeah they should pay for that in the form of a tax".

[cyberdemon]

Er, no, sorry but I don't want the government selling my personal data either. It's MY data, they should pay ME for it if anyone

[Anonymous Coward]

Yeah but your data might be worth less than mine, you might be absolutely worthless in the grand scheme of things as an entry in a database. It's better as a tax because then everyone can benefit. Also, as a tax it can be legally enforceable with some massive fucking teeth. You as an individual moaning about your £20 worth of personal information aren't worth a wank. which is why they get away with it. You fight scale with scale...if the entire population of a country taxes someone like Facebook etc for slurping data, they might sit up and take notice...if you as an individual demand a few quid for some lines in a database, you'll never be taken seriously. It's you asking for peanuts from a multi-billion pound corporation...they won't even fucking hear you.

I personally, don't give a shit what my data is worth, I just don't want them to have it...and if they want it, they can fix the fucking potholes in front of my house. I can't, as the typical owner of essentially fuck all value in the grand scheme of things, take Facebook to court and sue them for a few hundred quid...it just isn't worth it, the lawyers win...common sense dictates that if the entire country is opted in to fucking Facebook and putting the proceeds into..hopefully..making the country better then that's a better choice.

[cyberdemon]

> I personally, don't give a shit what my data is worth, I just don't want them to have it.

Well, that's what the 'if anybody' was for.

I.e. i wasn't implying that I or anyone else should sell their data, just that they should have a choice, and "not for sale" must be a valid choice

But the idea of the government selling our personal data from under us via a tax for using it, is repugnant to me

[Anonymous Coward]

It is repugnant, and I was clear that I don't want anyone to have my data under any circumstances...but the only way to dissuade companies from collecting data is to make sure it isn't financially viable...the only tool that really exists for that is taxes. It's what a lot of taxes are for. We don't tax fresh fruit and veg because we want people to purchase and consume that...it's good for you...but we tax cigarettes and booze because we'd rather people didn't do that. The freedom to do it is there, but it'll cost you.

The same thing needs to apply to data collection. We have to be honest and realise that entirely avoiding data collection is basically impossible now unless you want to remove yourself from society and go and live in a cave like a hermit...therefore a tax based on profits for companies that collect personal data makes sense.

Just because you or I are against mass data harvesting, it doesn't mean that the vast majority of people even understand what it is or that it exists. For quite a few people, just installing uBlock and Privacy Badger on Firefox are massive eye openers...until they can visually see how often tracking links etc are blocked they have no idea of the scale...nor do they care...they have to see it to understand it and for large swathes of the population, they will never see it...therefore, it's unlikely you're ever going to have a critical mass of people wisely standing up as individuals demanding that the tracking and collection stops...we have to go further than that...the government taxing organisations that collect data is pretty much the only way...the government won't even be handling our data, they don't somehow become our data broker...they simply enforce the tax. Sharing the data is still our choice...just the tax being in existence would massively increase awareness...public figures showing tax receipts and from whom would give the public some idea of who is collecting data and at what scale...and crucially, who is profiting the most. We already have those answers, we're part of the industry (ironically, a lot of us earn what we earn because we're being paid with money that was probably at some point the result of data collection in some form or other)...but Aunt Dorothy down in Bournemouth hasn't got a clue...we just look like tinfoil hatters to a lot of people...there are also people out there that are aware of the tracking and simply don't care because they get access to lots of free stuff...for them it's worth it.

The people out there, like us, that are staunchly against tracking and data collection make up a relatively small (albeit fairly vocal) group...but there aren't enough of us...therefore, we may have to accept something that is a bit stinky to get rid of something that is extremely stinky.

Us protesting against data harvesting and collection is background noise because if these companies can't collect our data, it's irrelevant, there are millions of others who don't care, don't understand or simply are unaware...so our individual data is a drop in the ocean and these companies don't have to give a shit.

[Anonymous Coward]

The government wouldn't be selling it, you would ultimately be in control of it...just the government enforces the rules and taxes companies for slurping it. The downside is most free services would disappear and your cost of living would go up. Significantly.

If this isn't palatable, there is always self hosting available...which is probably what most of the folks here do to some extent...that is also expensive.

Essentially, beggars can't be choosers...unless they get together as a massive collection of millions of beggars and demand their government regulates and taxes the behavior they don't approve of. Which will never happen because Joe Public generally doesn't care about their data being slurped as long as they're getting to use something for free. Mostly because they don't understand the value of their data, but also because quite a few people recognise that their data is actually quite worthless in the grand scheme of things. It's only worth something as part of a larger statistical dataset. Standalone, it's worth nothing.

[Anonymous Coward]

Pears are sweeter and softer than Apples, but apples are crispier and tastier, and they both grow on trees. So pears are like apples, but pears are better because they're sweeter.

Why are you comparing human intelligence to artificial intelligence? Who said the goal of artificial intelligence was to mimic humans? All we know from your statement is that AI is bad at mimicking humans...that doesn't make it bad...it's just another form of intelligence.

"when humans doubt the answers they work again with other people to learn about their doubts"

AI can do this as well...but people seem to skip over that part for some reason. Everyone seems to use one chat bot at a time and tar the entirety of AI with that one experience. That's not how I use AI to develop, I might use 3 or 4 different models in tandem, each one with a different set of pros and cons...if you're still asking chatbots to "write me python script that can loop through a CSV and put it in a JSON object" you're about 3 years behind...because that's not where we're at with it now...we're far beyond that.

ChatGPT (and <insert you favourite alternative chat bot>) is one example of an implementation of AI, it is not the AI in and of itself.

The reality is, I can go to something like Bolt (either .new if I want to use a cloud service or .diy if I want to work locally) or v0 or some equivalent, talk to it for a bit and have it build me a rough prototype in under 10 minutes for less than 20p in tokens, some places are even cheaper now thanks to Deepseek. I can then feedback some additional prompts to refine it to a point I'm happy with, I can then export the prototype and run it through a reasoning model, along with the original context and conversation and have it suggest improvements and refinements that I can then pass on to something like Claude which will improve the prototype. By this point I've still spent less than 2 hours and £1 to get to a compelling working prototype that I can demonstrate, gather human feedback on, then pass that feedback back in again to refine even further. Wash rinse repeat and for less than a tenner and a couple of days shepherding bots, followed by a manual code review (by me) I can have a product ready to go. Just a few years ago, that was impossible...I'd be paying hundreds of pounds, possibly thousands and spending weeks dealing with developers, explaining concepts, writing specifications, drawing diagrams, trying to convince other humans that think they know better etc etc arguing about the stack, dealing with people that don't call back or aren't motivated etc etc...

Is the resulting code better than a human can produce? That's up for debate...but is the experience better than working with a team of opinionated humans through a language barrier? You bet your ass it is.

Since the beginning of LLMs, my ability to prototype quickly and efficiently has gone through the roof. I have to acknowledge that AI as it stands is a step forward. I can't ignore it. I don't even need to be motivated myself, I can relax while the AI does the heavy lifting, I can sit there with a beer and just dream shit up.

Of course everything needs to be checked, but that's no different to hiring a bunch of other developers, I'd check their code thoroughly as well...the only difference is the AI gets the code to me sooner and can iterate faster. It can document it (good luck getting humans to do that most of the time), it will comment it and it will explain the code to a certain extent, which makes checking the code that much easier. I can go through several revisions in under an hour, let alone a day or even weeks.

You can be angry at AI all you like and quibble over the finer details over whether it is actually AI...but it doesn't matter, it's irrelevant...the quibbling and nit picking proves exactly why AI is generally better than humans. You no longer need a team of humans that will argue amongst themselves over "the stack" or "the process" or any other poxy opinionated detail that doesn't matter...you, as an individual, can just crack on and get stuff done, rightly or wrongly, the way you see it without any interference.

Over the span of my career, the two main barriers for me getting products built and out there is time and money, I've been capable of writing software since I was a young teenager and I've been writing code for 25+ years, ability has never been a barrier for me...before AI, a prototype might have taken me a few weeks to knock together into a demonstrable state, at which point I'd start looking for some extra hands to refine it. The time from conception to actually having some semblance of a final product could be months...that's a lot of time, and if you hire a team to work with you, that's also a lot of money...usually, one of these would kill the project dead. I'd either not have enough money or not enough time or both.

My problem now, is which project do I launch first? I can have half a dozen "finished" products in a week in a state that is at the very least "beta" quality that I can distribute to friends and family for testing and feedback etc.

You just can't ignore this...I'm sorry, but if you're still hand coding things and working exclusively with humans you're trying to win a war against tanks with bows and arrows...the biggest danger in ignoring and writing off AI as it is, is that you might shop around for some developers to help you, and when they turn you down, they can have a prototype cranked out by AI and beat you, before you've even finished hand coding your alpha.

By all means though, stick your fingers in your ears and scream "la la la la" all you like...by the time you take your fingers out of your ears and open your eyes, you'll be folding Happy Meal boxes and stuffing them with nuggets and toys...mark my words.

There is nothing wrong with being skeptical about AI, despite the benefits I get from it, I am still skeptical about certain things...and it is clear that sometimes it does produce absolute garbage...but entirely dismissing it, that is career suicide at this point...maybe not immediate career suicide, but at some point, you're going to find yourself to be the anachronistic element in a given team.

[amanfromMars 1]

Re: Better Sweeter Apple-like Pears @AC

11/10 for that upvote worthy post, AC. :-) The future is bright, the future is spooky surreal action from a near distance ...... and aint that the Gospel to AI according to IT ‽ .

[Anonymous Coward]

Re: Better Sweeter Apple-like Pears @AC

Thanks. Totally agree.

[Anonymous Coward]

And then it comes down to the fine print and the conjecture of ownership.

Since you use AI to write your code, the platform now technically owns this code and, soon enough, this assertion will come true and all your work will not be owned by the AI and, therefore, the owner of the AI.

[Anonymous Coward]

No it doesn't. I use local LLMs for the final touches and clean up, I'm not using a "platform" for the production level stuff. It might sound outlandish to you, but owning a bunch of GPUs (which you can find relatively cheaply, if you know what you're looking for, that are capable of running larger models) is a thing. The demand for GPUs isn't just datacentres, it's also professionals.

There are models that permit commercial usage. This is what I mean by people not seeing the full picture. People think chat bots are the be all and end all of AI...they are not.

Professionals in tech, just like any other profession invest heavily in kitting themselves out and giving themselves an advantage. A professional techie is not a person with a waxed moustache in a coffee shop, listening to Coldplay, on a Macbook connecting back to his Mac mini at home. That is an amateur that thinks they're a professional.

For a professional, dropping £10k or more on some kit is nothing because that £10k investment will earn you 10x that back.

[Omnipresent]

especially when the decision are being made out of ignorance, and not knowledge.

ignorance is not a crime.

[Eclectic Man]

But, at least in theory, 'ignorance of the law is no defence'.

[Like a badger]

"ignorance is not a crime."

Bloody well should be.

[Anonymous Coward]

Failing to properly inform the ignorant should be a crime.

[nsld]

Re: Four simple rules

A lot of confusion exists on what AI is, for starters ordering milk when you're getting low is basic stock control maths, not an AI driven decision derived from a large language model.

A lot of 'AI' branded stuff is really just a glorified search engine, or an expert system following pre defined rules.

Until we define it properly it's difficult to apply any form of oversight to it. Especially with the number of things calling themselves AI to look cool and impress investors...

[sabroni]

Re: All knowledge should be free.

Says the poster hiding their handle.

[rg287]

Re: Four simple rules

1) Make the AI companies liable for the output of their models; treat them like publishers.

1) Won't work...because anything is a weapon in the wrong hands. If you started to try and blame a DIY store for all the bad stuff that might happen if they sell someone a hammer, all that will happen is that DIY stores won't sell hammers, or worse they will sell "safety hammers" made out of cotton and fluff which can't hammer a nail...which punishes the majority of people that use hammers correctly. The penalty for AI misuse should land squarely on the user not the AI.

If you sell your AI as "suitable for giving legal advice" then the AI firm inherits a level of responsibility for what their customers do with it - this is some combination of "Know Your Customer" and "Trade Descriptions Act".

It's not just the Post Office getting it in the neck for wrongly prosecuting people based on Horizon outputs - it was Fujitsu who aided and abetted, insisting their system wasn't a crock of shit. There is no reason why Anthropic/OpenAI/whoever should not bear some responsibilty as the original vendor to ensure their product is fit for purpose and - to the best of their ability - is not used inappropriately. They could easily include Terms of Service to the effect "This model must not be used to make unsupervised decisions about humans. Where used for critical-to-life decisions, a qualified human must be remain in the loop". Only with better wording.

Somewhat agree, but it depends on the decisions the AI is making and what it is they control...nuclear silos...hell no...ordering milk when I run out in the fridge, why not?

Why would you need an AI to reorder milk?

A simple

if 'milk' NOT IN fridge.contents:

order('milk')

would suffice.

You could even have a small statistical learning model that learns your average reorder rate and can order milk a day in advance of your likely running out based on an average burn rate, or orders extra milk because you use more over the weekend (because you like porridge on a Sunday morning).

None of which entails a full-fledged AI model. It's basic stock-control mathematics that people have been doing for hundreds of years.

There are likely decisions that humans are unable to make for ideological reasons or refuse to make for malicious reasons that an AI can make without any regard for ideology and without malice that ccould be generally beneficial, until we test it, we don't actually know though...decision making based on facts without distortion is potentially one of the key benefits of AI

Show me an unbiased set of training data and I'll show you a goose that lays golden eggs.

All knowledge should be free...even to students. Personal information, yeah they should pay for that in the form of a tax.

I can get on board with that for academic publishing - since the researchers have already paid for the research and to publish, the publishers should be able to basically give it away for free. But are you saying that all books should be produced for free(!), given away to libraries and bookshops on the basis of "help yourselves".

I suppose we could treat fiction separate to non-fiction, but people have literally some right to be paid for their work. There is a balance between "no such thing as copyright and I can sell hooky copies of Pratchett at will", and the exploitative academic publishing racket that we curently live with.

[Jimmy2Cows]

if 'milk' NOT IN fridge.contents: order('milk')

Hmm... that statement places a milk order every time I take the milk out of the fridge to, you know, use it. Not convinced you've thought this through.

You at least need a minimum time window to expire where milk is not in the fridge before ordering more. And I don't want it to wait until there's no milk before ordering more. It needs to see I'm running low, but not out, and then order in advance so I don't run out. Then it needs to adjust for different milk bottle sizes, different types of milk (full, semi-skimmed, skimmed etc).

Does it need AI? Maybe it needs image recongition ML to evaluate milk bottle size, milk type, and how much milk is left. Does it need an LLM for that? Hell no. Wrong sort of AI.

[elsergiovolador]

Re: Four simple rules

That's some sort of cyber-nimbyism.

Are you okay?

[Anonymous Coward]

Re: Four simple rules

Not really, I'm a big advocate for AI and leaving it alone for now...there is little point in hobbling it before it's got up off the ground...we have existing laws and frameworks that clearly set out the SOP of being in society...simply hold the user responsible.

If someone commits murder and uses AI to do it, they committed murder...it's as simple as that.

If someone commits murder and you have a long drawn out debate over whether it was actually the AI and the human had little to do with it, lets redefine murder...is a very American "loop hole your way out of shit" approach...ah well it's slightly different therefore it might be completely different, lets get a bunch of fuckwits, nutters and so called experts into a hearing and debate that.

Right now, we're so early with the tech, that I don't think many people can genuinely call themselves "experts" at this point and keep a straight face. We need to allow room for experts to actually exist before we can debate anything, let alone understand what the fuck we have...we can't assess the impact of something until it has had some kind of significant impact...it's like analysing a crime scene before a crime has actually happened. Might be a murder here at some point in the future, lets dust for prints now to save some time.

[Mentat74]

So it's OK then...

For this so-called 'A.I.' to keep spouting bullcrap as long as it's not 'criminal' bullcrap ?

Defamation is still OK as long as it's not politicians, rich or famous people that are being defamed ?

[Anonymous Coward]

Re: So it's OK then...

It's wild to me that people think the extent of AI is just "free" online chat bots that summarise things...it really, really isn't.

There are applications of AI that exist today, right now, and have done for quite some time that are provably better than humans at various things. LLMs like ChatGPT are not the pinnacle of AI technology...what we get to access via chat bots does not represent the full capabilities of AI...or even LLMs...what we get to see through the use of chatbots like ChatGPT, Gemma etc are lower parameter models...the higher parameter stuff which is much more capable is still locked behind mountains of money and expense.

Essentially, what you can do with AI right now isn't a patch on what someone with considerable resources can do with AI...and that financial barrier to entry has dropped orders of magnitude with Deepseek.

We are so early with technology like LLMs, it's unreal...

[ChrisElvidge]

Re: So it's OK then...

"LLMs like ChatGPT are not the pinnacle of AI technology"

They're not even AI

[Anonymous Coward]

Re: So it's OK then...

Well, they are. Just not particularly well executed AI...the mathematical and technical underpinnings are all there...in fact the maths has been around for nearly 75 years at this point...LLMs are not based on some new concept that only popped into existence 5 years ago...it took time for technology to catch up and become cheap enough in order for it to become feasible.

[Omnipresent]

Re: So it's OK then...

I believe it was on this very tech rag there was a story not long ago, about a military general saying he personally fought an f-16 controlled by A.I. and lost.

[Anonymous Coward]

Re: So it's OK then...

Well you'd expect that. Senior officers don't (or shouldn't) fly fast jets, as sadly that's for the younger generation with faster reflexes.

[rg287]

Re: So it's OK then...

It's wild to me that people think the extent of AI is just "free" online chat bots that summarise things...it really, really isn't.

So, here's the thing... most people here know that. They're concerned that politicians don't. Or that politicians think that a sophisticated pattern-matching system constitutes "Intelligence". Or that the best way to run some sort of analysis is to feed data into ChatGPT - when in reality even if an ML model could be the most appropriate analysis tool, you don't want a f-ing LLM.

There are lots of genuinely useful Machine Learning applications lurking under the hype-train umbrella of "AI". Medical imaging analysis, an adjunct to existing fraud detection/forensic tools, environmental & habitat monitoring, rail infrastructure condition monitoring. All sorts of things where you're sifting large datasets with some sort of Statistical Learning methodology.

But the political take-home from that seems to be that we need to give AWS/CoreWeave/whoever a carte blanche to ignore planning regs, throw up DCs wherever the f--k they like and generally subsidise/legitimise a bunch of billionaires. All this under a banner of "securing British AI expertise" when we're actually just subsidising a bunch of Data Centres so that the calculations are performed within the borders of the UK.

If we want to innovate in this space, the sensible thing would be to develop a national facility (much like Edinburgh Uni's exascale super computer that Labour cancelled funding for last year), or help universities develop small clusters (£6-10m each) and then sponsor a bunch of PhD projects into those genuinely useful applications like medical analysis. That's how you create a domestic pool of expertise.

There's no need to be sucking off Sam Altman or Peter Thiel to "attract private investment" or pay Palantir to do this for us. We have what we need, just leverage it.

[Jonathon Green]

It’s a matter of spelling…

How long do we give it before Claude becomes clawed? :-)

[Anonymous Coward]

Re: It’s a matter of spelling…

Probably ages, because Anthropic can't even figure out how to implement a working payment system...the barrier to entry for Claude isn't that it is massively complex to use or even difficult to understand...the barrier to entry right now is their garbage payment gateway setup.

[elsergiovolador]

Re: It’s a matter of spelling…

When we will get Benedict?

Claude just can't stop talking about baguettes and onions, which is tiresome.

[Anonymous Coward]

Re: It’s a matter of spelling…

but Benedict won't stop talking about eggs ...

[Anonymous Coward]

Re: It’s a matter of spelling…

Just don't ask GlasgowGPT about deep fried mars bars

[mark l 2]

How does this tie in with the online safety act where harmful but not illegal content has to be removed or face big fines?

Since it wouldn't surprise me if all these LLM with the right questions are able to provide details on how to do harmful things such as dangerous weight loss techniques, suicide advice or information other bits of harmful information that are not outright illegal.

[Dan 55]

Peace for our time

Silicon Valley gets to keep the bubble going a bit longer, the UK's society enshittifies a bit more thanks to AI as it did with social networks, and the US government will bank the win and carry on with its other idiotic demands.

[Wang Cores]

Re: Peace for our time

It will go down in history as similar.

[Handlebars]

local council on the front foot!

Swindon council making documents more accessible is nice to hear.

[Dan 55]

Re: local council on the front foot!

Hopefully it just changes the formatting and not the content.

[IanRS]

Re: local council on the front foot!

Not sure it needs an AI to do that. Surely just swapping what CSS template is applied to the page could do that, which might even lead to the content not being rewritten to mean something different.

[PRR]

Re: local council on the front foot!

> Surely just swapping what CSS template is applied to the page could do that, which might even lead to the content not being rewritten to mean something different.

True only if the CSS has been consistently applied. When does that happen?

> documents of 5-10 pages cost around £600 to convert

So ten hours to format a page? Really? HTH did my mother lay-out entire 60-page newspapers (minus ads and funnies) with images in a 6 hour day?

[doublelayer]

Re: local council on the front foot!

It is, but I have no idea how they were doing it before.

"The result has been significant financial savings, it's claimed. Where previously documents of 5-10 pages cost around £600 to convert, Simply Readable does the job for just 7-10 pence, freeing funds for other social services."

If reformatting a document costs that much, I'd like to have the customer list for whoever was doing it before. You don't need an LLM or a lot of expense to change the font size and spacing and scale up some images. While there are probably some documents that needed more than that, I have a feeling it wasn't most of them, and even those should not cost anything near that much to redo.

[veti]

Re: local council on the front foot!

"This staggering figure underscores the transformative potential of " - narrowly defining a specific problem for your tool to solve, not trying to make it do everything at once.

When more people learn to use AI that way, I expect great things of it. Sadly, this whole idea of "general AI" is blinding everyone to the true potential. Just because its training and data and scope is "general", doesn't mean it's the best way to use it.

[Omnipresent]

This discourse is good and right.

I like what I see and hear from the UK on this matter. This is necessary discourse that was ignored in the case of the US.

All that I have for you is this.

These are not business decisions. That was the mistake America made. Do not use America as a model, America no longer exists. Consider these early discussions war time preparations. War is coming to your shores (if it hasn't already), and it will be tested early and often.

As far as safety regulations are concerned, I would insist on a multitiered approach. Have your ethos, and your regulations, but make them adaptable. War changes. You will need to turn these protections on and off as needed. I would suggest a small group of responsible decision makers. Too big, and bureaucracy makes it useless, too small, and power becomes consolidated to a few.

I would also make my ethos one grounded in an innate knowledge of right and wrong, good and evil. A recognition that there is something bigger than yourself. You are laying a foundation for future generations, America failed.

Lastly, remove JD vance and friends from your shores immediately, and never let them step foot on your shores again. You are already being manipulated.

[Eclectic Man]

Re: This discourse is good and right.

Omnipresent I would also make my ethos one grounded in an innate knowledge of right and wrong, good and evil.

This is actually rather difficult. In every major war to minor skirmish in human history, each side considered themselves to be 'on the right side', or 'the good guys'. Having been brought up in a mostly Christian, Western country, I have some sense of 'good' and 'evil' but I cannot create a definition of either of them that serves to characterise them well enough that anyone else could determine exactly what is or is not, either of them. I can come up with lots of specific examples of actions that I consider to have been 'evil' or 'wrong', or 'good' or 'right', but definitions are tricky and what an AI actually needs. The training data could easily include examples for an AI to work with, but that is not the same as an axiomatic definition.

J D Vance is an interesting case. His recent speech at the security conference in Munich was allegedly mostly, if not entirely, for his domestic USA audience. There is a response to some of his allegations in the BBC web site: https://www.bbc.co.uk/news/articles/c4g9nmeyzkjo . As for removing him from our shores, he is the democratically elected Vice President of the USA and it is not feasible to exclude him. Oh, and if you think that we 'only' have four years of this administration, my guess is that he has set his sights on being President 48. But regarding free speech, remember it was Oliver Wendell Holmes who stated that "freedom of speech does not give you the right to shout "fire!" in a crowded theatre". (https://scholarship.law.wm.edu/cgi/viewcontent.cgi?article=1748&context=wmborj )

[veti]

Re: This discourse is good and right.

I can come up with lots of specific examples of actions that I consider to have been 'evil' or 'wrong', or 'good' or 'right', but definitions are tricky and what an AI actually needs

No, an AI can manage quite well without "definitions" of the sort that philosophers have forever been quibbling about. Whether you'd agree with its assessments, that's another question. The key point is that both the AI itself, and the people overseeing it, should be able to agree that there is such a thing as good and evil, or right and wrong, and it is sometimes worth sacrificing functionality to put yourself on the right side of that divide.

I actually think Vance had a point. (Not much of one, and what there was, was gravely undermined by his own refusal to accept the result of the 2020 US election. But there is a case for Europe to answer, it has not been doing particularly well of late.) But I don't think he'll be the 48th president - I don't think there's any plausible career path in elected office after being Trump's VP.

[rg287]

Re: This discourse is good and right.

His recent speech at the security conference in Munich was allegedly mostly, if not entirely, for his domestic USA audience.

Given how well it went down with the audience in the room... I'd say entirely.

As for removing him from our shores, he is the democratically elected Vice President of the USA and it is not feasible to exclude him.

It most certainly is. We're literally under no obligation to entertain someone who is going to decry "misinformation" and then minutes later go on a screed about how Scots have been warned against private prayer in their own homes (they haven't).

The most impressive part of his whole speech was that noone shouted out halfway through "What the actual fuck are you talking about you incoherent halfwit?". I certainly would have - perhaps why I am not a diplomat! The universal response from journalists present was one of stunned "WTAF did we just sit through? This is a security conference isn't it?"

Now, some will say that we actually can't exclude Vance, because that would effectively break up NATO by virtue of Europe telling the US to do one. However, it seems likely that the US is well on the way to "doing one" all on their own... in which case we can afford to limit our tolerance of their misjudged leadership. Don't let the door hit you on the way out. Of course we have nothing against America in principle - the current administration are just a bit hard to work with, and we welcome the next lot.

Meanwhile, if Russia wants to try anything, we'll see how well their air force fares against the Rapier and Sky Sabre installations dotted along Europe's side of the border. This new era of drone warfare could get very unpleasant for civilians in the border areas of Estonia/Latvia, or in Poland (if they came through Belarus), but Russia has shown that it's basically not got the personnel or materiel to come rolling over the border Cold War style, and they know it. If Vlad goes for the big red button, the US can't help us anyway.

With no ability to wage conventional warfare, the biggest threat from Russia is actual misinformation campaigns - sowing the seeds of the sort of social division that put him in power.

[codejunky]

Re: This discourse is good and right.

@rg287

"It most certainly is. We're literally under no obligation to entertain someone who is going to decry "misinformation" and then minutes later go on a screed about how Scots have been warned against private prayer in their own homes (they haven't)."

I wonder if his comments were in relation to the reporting last year about such rules-

https://www.gbnews.com/news/praying-at-home-illegal-scottland-safe-access-zones-abortion-law

While it might be a misinterpretation of what the Scottish government means I dont know, it does seem to have got Christian sites jumping all over it last year so maybe that is where Vance heard it from.

[rg287]

Re: This discourse is good and right.

https://www.gbnews.com/news/praying-at-home-illegal-scottland-safe-access-zones-abortion-law

I can see where he's going wrong if he's using GBNews as a source for UK Affairs. An Emiratii-controlled network that serves to divide opinion and sow discord. And leads the UK news community only in the chart for upheld complaints and sanctions from OfCom for saying things that are known to be untrue.

Can be safely treated the same as RT.

[codejunky]

Re: This discourse is good and right.

@rg287

"I can see where he's going wrong if he's using GBNews as a source for UK Affairs."

It wasnt just GB news but various Christian outlets too which is probably where he heard it from. Also it appears some homes did get letters warning them they are within such zones and the law. I assume it comes down to the police interpretation and enforcement but reading the wording it does seem to suggest just not being seen praying for the dead. From a quick search I found this kind of reporting in Oct 24 but nothing that seemed to refute it.

It doesnt detract from his point about Europe and the last 4 years of the US so it is nice to see someone recognising the issues. The UK has just demanded the right to access apple users accounts- https://www.bbc.co.uk/news/articles/c20g288yldko

[ChodeMonkey]

Re: This discourse is good and right.

"While it might be a misinterpretation of what the Scottish government means I dont know, it does seem to have got Christian sites jumping all over it last year so maybe that is where Vance heard it from."

Very nice post. "might be a misinterpretation", "I dont know", "maybe that is where Vance heard".

So bringing this snippet in as an argument, but then rolling it back. But not acknowledging that it really is poppycock.

A masterclass. Thank you Ma'am.

[graemep]

Re: This discourse is good and right.

> then minutes later go on a screed about how Scots have been warned against private prayer in their own homes (they haven't).

True, it is a significant exaggeration, but people have been arrested for silent prayer - it is legal to stand silently by the side of the road, but it is illegal in certain places to stand silently by the side of the road praying! Its also possible it would be illegal to pray on your own property outside your house (e.g. in your front garden), I think.

The online safety act (another things the Trump administration are reportedly pressuring the UK about) is another disaster with things like forums closing - see earlier articles here.

Behaving in a way that makes what Vance says partly true helps him both domestically and internationally.

[ChodeMonkey]

Re: This discourse is good and right.

"people have been arrested for silent prayer"

Have they?! When? Were?

[sabroni]

Re: and never let them step foot on your shores again.

Arrgh! Every time i hear ear that phrase it drives me up the fucking wall. I have to punch fist the wall and run leg it to a dark room to sit bum down and cry eyes.

You don't need the word foot when you say step. You can't step with anything else.

The phrase you are looking for is "set foot".

Yes, I do have a lot of time on my hands today, how can you tell?

[amanfromMars 1]

For Nations Speaking Peace unto Nations, an AID for Revolutionary Progress

As AI continues to rapidly and rabidly develop, you can expect it to be increasingly effective in keeping countries safe from government policymakers and all similar wannabe Caesars/deluded wishful thinkers via a vast and expanding networking array of almighty stealthy and surreal untouchable means with every kind of novel remote virtually controlled and strange strangling and entangling memes.

Quite whether the current UKGBNI package is equipped or even able and enabled to be able to equip itself with the intelligence streams from A.N.Others needed to not fall foul of, and into competition against or opposition to such an third party AIDevelopment, is something which all too soon becomes perfectly apparent ........ with every indication being that it will be painfully so.

Donkeys leading lions always results in such evidence.

[Omnipresent]

Re: For Nations Speaking Peace unto Nations, an AID for Revolutionary Progress

"A.N.Others needed to not fall foul of, and into competition against or opposition to such an third party AIDevelopment"

Anthropic has the opportunity to be better. Other A.I.'s have fallen into the wrong hands. There may be a time sooner, rather than later, that these A.I.'s are set against each other in combat. One will rely upon it solely, the other will take control of its technology.

[amanfromMars 1]

Re: For Nations Speaking Peace unto Nations, an AID for Revolutionary Progress

Anthropic has the opportunity to be better. .... Omnipresent

All have the opportunity, Omnipresent. And the best of the best will be the One and the Singularity destined to rule almightily and reign overwhelmingly over any competition and all opposition.

The only presumably troubling question, to many with worries enough and aplenty in present inherited Establishment systems and remaining to be answered, is whether it is to be a leading wild wacky western delight or a pioneering exotic erotic eastern confection. ...... an Alphabetic DeepMind or High-Flyer DeepSeek clone/drone/LOVEchild/bastard?

Or neither of the two floated above?

[Omnipresent]

Re: For Nations Speaking Peace unto Nations, an AID for Revolutionary Progress

When you put power in the hands of a few, it's called a dictatorship. When you have an over abundance of opinions, it's called a mess. A mess is easily manipulated to a few.

[ComicalEngineer]

Ah, the UK government -- decisions on IT and AI made by someone with a degree in Politics, Philosophy and Economics, advised by a *consultant* from big business.

[Eclectic Man]

Umm, do you mean:

"Parliamentary Under-Secretary of State for AI and Digital Government

Feryal Clark MP

Feryal Clark was appointed Parliamentary Under-Secretary of State at the Department for Science, Innovation and Technology on Wednesday 9 July 2024. She was elected Labour MP for Enfield North in July 2024.

She studied Bioinformatics at the University of Exeter. "

From:

https://www.gov.uk/government/ministers/parliamentary-under-secretary-of-state-for-ai-and-digital-government#:~:text=Current%20role%20holder-,Feryal%20Clark%20MP,on%20Wednesday%209%20July%202024.

[Steve Davies 3]

Wow, a gubbermint minister underling who

is not a PPE grad.

Are you sure that her CV is right? After all, the one that the Chancellor gave out is very economical with the truth.

Either way, that degree does not mean that she has any idea about deploying AI on a national scale.

[Eclectic Man]

Re: Wow, a gubbermint minister underling who

The web site says she studied bioinformatics, at Exeter not that she actually got a degree in it.

I was genuinely asking whether the OP meant her.

[Roj Blake]

Remember when Kwasi Kwarteng was Chancellor? He's got a doctorate in Economics, so presumably must have done a great job?

[Brewster's Angle Grinder]

"No details are offered on whether this AI savings entailed a cost in jobs or expenditures in the form of Jobseeker's Allowance."

First, you are assuming they were fired rather than there time used for other things.

But if they were, I guarantee UC/JSA was far less than they were earning.

[Tron]

AI is a perfect fit for local and national government in the UK.

Widely despised, untrustworthy and prone to failure.

One concern: Are these chatbots capable of contracting for services with Oracle?

Incidentally, 'Unless it's causing serious bother, you can crack on' is not a bad plan for tech. We don't want to go down the road of locking people up or banning tech for what they/it might possibly do. You should have to do something criminal before the old bill come after you.

[doublelayer]

Re: AI is a perfect fit for local and national government in the UK.

Sometimes, it is helpful to define what serious bother would be first and forbid it, then give them free rein in what remains. Otherwise, they will create plenty of serious bother, they'll just think that you gave them permission to do so first and consider cleaning it up someone else's problem, that is if they ever stop doing it. This is how we ended up with lots of data collection requiring repeated patches, the largest of which, GDPR, is not enforced because the people concerned have been abusing so much data over the years that we can't exactly make them stop without consequences*. It might have been better if we ruled that out at the beginning, and businesses could have found different ways of operating that would have complied with it.

* I'm willing to accept those consequences, since it's mostly things like Facebook losing lots of money. However, my willingness isn't enough to make it happen.

[sabroni]

Re: Are these chatbots capable of contracting for services with Oracle?

Quality!

[Vader]

We have aligned

We have aligned with the US for AI. TBH the USA has vast AI systems what will the UK contribute really, it will be peanuts. As the US has recently shown allies means nothing, this special relationship is in the mind of all PM's, but what does it actually mean, it's all US sided.

As far as Trump and Co ae concerned it' Make America Gay Again(MAGA).

[Anonymous Coward]

The current UK Government

like most of those since Harold (pipe and Gannex) Wilson Mk 1, cancelled the TSR2 (causing my family to have to move house again) do not have a clue when it comes to technology.

The current shambles in power demonstrated that with their approach to AI. That horse had disappeared over the horizon before they woke up from their post lunch siesta.

The phrase "I'm sorry, I don't have a clue" is spot on. That is followed by "We must be seen to be doing something," (the Captain Mainwaring response) when it should be the Cpl Jones "Don't Panic" shouted out loud.

We are Doomed I tell you, Doomed.

[amanfromMars 1]

Re: The current UK Government

And what is equally galling and revealing and mystifying, AC, is the mainstream media’s constant supportive presentation of their piffle and waffle which only encourages their pathetic flights of fancy and self-delusion to be tendered as a matter of future activity to enliven and reflect secretive closed shop policy ambitions/targets ........ in vainglorious attempts to grow the economy with nothing to share and everything in a dire straits state of despair and disrepair

And that indicates a monumental lack of smarter intelligence in that vast swathe of humanity too ...... aiding and abetting the rapidly approaching incoming Doom you foretell.

I’d venture the UK is FCUKd ...... and the future news will demonstrably reflect it to be so for there is no leadership delivering anything novel or great to any man or beast anywhere.

[codejunky]

Again

Another benefit of brexit then to not align with the EU in regulating AI out of progressing.

[UnknownUnknown]

Re: Again

Sounds like another Unicorn of deregulated AI and Orange Jesus Tech Bro insanity to give a body we’ve too: esp. as the ROI is so poor.

“American Brexit”.

[codejunky]

Re: Again

@UnknownUnknown

"Sounds like another Unicorn of deregulated AI and Orange Jesus Tech Bro insanity to give a body we’ve too: esp. as the ROI is so poor."

What are you talking about? You seem to be jumping all over the place with your ranting sentence.

[LionelB]

Re: Again

Depends, I suppose, on the extent to which you equate "progress" with concentration of control over development and deployment of technology which affects all our lives in the hands of unelected and unacountable bottom-line-driven mega corporations run by power-hungry narcissists and grifters.

[UnknownUnknown]

How much ?!?

“The result has been significant financial savings, it's claimed. Where previously documents of 5-10 pages cost around £600 to convert, Simply Readable does the job for just 7-10 pence, freeing funds for other social services.

According to the UK's Local Government Association (LGA), the tool has delivered a 749,900 percent return on investment.

"This staggering figure underscores the transformative potential of 'Simply Readable' and AI-powered solutions in promoting social inclusion while achieving significant cost savings and improved operational efficiency," the LGA said earlier this month.”

To convert 5-10 pages into accessible format seems grossly inflated cost. Who were they paying Cr@pita ??? Feels like an hours work for any half competent Admin/PA.

I’m sure many of your existing staff would have done it on a T&M basis out of hours for a little OT….

£600 is insane.

[navarac]

UK Government...

The UK Government, like most governments, councils etc, are full of Nupties who seem to have been failures in other jobs. No wonder the world is in a mess.

[codejunky]

Re: UK Government...

@navarac

"The UK Government, like most governments, councils etc, are full of Nupties who seem to have been failures in other jobs. No wonder the world is in a mess."

That is the expectation of the skills of government. Add that they run on out of date information for too large an area for them to understand and the problem is letting them interfere too much.

[amanfromMars 1]

Re: UK Government... and their Responsibility for being Accountable for Civil Disobedience

The UK Government, like most governments, councils etc, are full of Nupties who seem to have been failures in other jobs. No wonder the world is in a mess. ..... navarac

Regarding the mess, what about the idiots that vote them into government office on the basis of their false party manifestos and broken and impossible to achieve electioneering promises and allow them to continue to plunder and destroy any semblance of national wealth and wisdom for a defaulted standard government term period of 5 very long years ... [tThe Fixed-term Parliaments Act (2011) passed during the Lib Dem-Tory coalition years mandated that elections be held every 5 years on the first Thursday in May. There was a caveat in place that a general election could be called if two-thirds of the House of Commons voted for an early general election or the government lost the confidence of the House.]

Methinks the chances of Parliamentary chancers calling for an early general election which sees them removed to the dole queue with zero likelihood of claiming sundry expenses is zero ..... and thus be their inaction tantamount to inciting rebellion and popular insurrection?