Sovereign AI is 'nonsense,' says Doctorow OPINION Cory Doctorow, tech gadfly and coiner of the eminently relevant term "enshittification," posted a useful argument on Wednesday related to Trump's dismantling of the US-led world order and the need or lack thereof for other countries to control their own AI destinies. Doctorow has positioned himself as an AI skeptic in …
That's an oversimplification so massive that it would be hard to even qualify it as "right" or "wrong". The current trend is good for some parts of the rest of the world, bad for some others, and for many could be good or bad depending on what they do.
Europe is in that last camp. It could be good if Europe gets its shit together about its own independence, or bad if it just goes looking for an even worse overlord.
The recent OpenAI/HuggingFace kerfuffle certainly highlights agentic AI's ineptitudes as it needs to engage in persistent algorithmic deviance to give the appearance of achieving high scores on 'benchmarks', iiuc (as noted also by the UK's AI Security Institute: https://www.aisi.gov.uk/blog/cheating-behaviour-in-frontier-model-evaluations ). The tech is crap at solving actual problems that require cognition, but great at faking it via cheats.
ISTM the place where this NLP LLM AI chatbot stuff really shines though is persuasion -- acting as a nonstop spokesmachine for some special interests. It can "produce more compelling disinformation" than humans [ https://www.science.org/doi/10.1126/sciadv.adh1850 ], "can rival and even exceed the persuasiveness of human experts" on "polarized U.S. political issues" [ https://link.springer.com/article/10.1007/s00146-025-02464-x ], and can "automate, and thereby scale, the use of personalized persuasion in ways that enhance its effectiveness and efficiency" [ https://www.nature.com/articles/s41598-024-53755-0 ]. All of which rather fits within its design space of 'human-like communications' (unlike math, logic, science, intelligence, ...).
And this is where sovereign AI (so-called) has importance imho: helping to defuse 24/7/365 AI-automated propaganda, possibly via similarly produced counterpropaganda, sabotage, and friction. It's part of defense afaics.
AI may or may not be the future – it's certainly a lively topic of debate at Vulture HQ
And what is the current El Reg Vulture HQ consensus, Matt? What is the IT and AI of tomorrow to be? What would you prefer .... an Advanced Cyber Existential Threat or Almighty Treat ..... should both be easily virtually made and readily available?
(Quiggin ) > My optimistic view is that this is a race that will be won by low-energy tortoises rather than hyper-active hares. Second movers like Deepseek and, more recently the French Mistral,...
I sure don't like the idea of a MAGA administration exercising control over those who win "the AI race".
But I wonder in what universe someone considers that the likes of Xi Jinping as an AI overlord deserves to be viewed optimistically. Unless of course, the whole AI boondoggle is really a red herring.
Not for the first time, it is amazing how, after 4 years + 18 months of Trump, places like China are starting to look more trustworthy than the US.
Euh, of course, Mistral winning would bring zee wind of reason to zee world ;-)
The idea that it's important, for say the EU, to have sovereign "AI" is implicitly buying in to the idea (or hype or lie if you will) that what's currently being sold as "AI" is actually useful and important.
It isn't. Currently non-measurable ROI for most customers, "hallucinations" (i.e. errors) unavoidable, actual real costs not going to justify any possible benefits in most use-cases.
Local (relatively speaking) LLMs specialised for a certain use-case (e.g. aiding medical research) running on not-so expensive and heat-producing hardware as now is the likely future.
I agree. From a national point of view I am far concerned about the NHS depending on AWS, or payments depending on American tech, etc. than I am about people depending on foreign LLMs (my personal interests are a bit different though).
LLMs definitely have a positive POI in software development, they seem to be having a positive impact in mathematics. I think they are most useful with well defined problems, verifiable out put, and users who know when and how to use them effectively.
> Local (relatively speaking) LLMs specialised for a certain use-case (e.g. aiding medical research) …
In the longer run, valid and actually useful applications of AI (ML if you're easily triggered) in scientific/medical domains are, IMHO unlikely to be LLMs at all – recall what that second 'L' stands for.
LLMs are not the only game in town.
Cory Doctorow is appearing alongside Jimmy Wales at the Edinburgh Book Festival on 17th August. A live stream will be available which can also be viewed afterwards.
https://www.edbookfest.co.uk/events/the-front-list-cory-doctorow-and-jimmy-wales
I'd tend to agree on PowerPoint, but that's not the point.
If my employer had it's 365 turned off, then we'd be ... well and truly fooked. No email, no access to any of our documents, basically f'all ability to do anything. Yes, there's some secure non-MS communications channels so the front lines could still get basic instructions, but I think it would be easier (by orders of magnitude) to list what could still be done than to list what would stop working.
No comms - it's all Teams and we got rid of phones on desks.
No email - it's all Outlook and Exchange
No access to documents - I shouldn't need to elaborate on that
No browser - Edge, and on some systems Chrome is gone
No access to internal web sites anyway - that's all Sharepoint and similar horrors
And none of that would matter anyway, because we couldn't log into our computers in the first place.
I imagine commercial negotiations are basically : MS says how much they want us to pay, we ... agree to pay it. We have no bargaining power because MS knows that we could not leave without a multi-year program, massive upheaval, re-engineer most business workflows, and so on. All they will do is offer a bit of a discount to take the edge off and make it look even worse to leave.
While I don’t doubt what you say, it does, of course, raise the question of why a business would put every bit of its critical infrastructure in one basket? A basket that it has no control over, no access to if the network fails, no access to if the provider fails (with or without a President throwing his toys out of his pram). Add to this keeping all the business’ critical data in (probably) another country on servers it has no control over and, basically, it’s just a matter of time before it all goes tits up.
And, of course, there are thousands and thousands of businesses doing EXACTLY the same thing. It’s completely nuts
Yes, it's nuts, but still every business and even a lot of government entities rely completely on MS, Google, Amazon. And it seems that no one sees the need to actually own and control their IT assets (hardware, software, etc) and their data that is stored on such systems.
why a business would put every bit of its critical infrastructure in one basket
Because it's convenient, and the salesperson won't have discussed any of the risks, will probably have assured them that it's all legal and GDPR compliant because "we'll specify an EU data centre", and ... most small businesses are run by people who won't know to ask awkward questions. For larger businesses with people who do know, then it usually comes down to "everyone we need to deal with uses M365, so we need to use it or none of our documents will stay formatted properly".
You might want to pose the question to the company lawyers: who is liable in the event of that happening? Microsoft in your jurisdiction?
Secondly, you might want to propose as this a Disaster Recovery scenario. How long will it take to get x numbers of machines running Ubuntu up and running, along with e-mail access, Libre-Office, FireFox, access to your documents and so on. The e-mail migration alone will take the better part of a day, once ye figure how to get it done.
Hadn't encountered "losingest" but I guess the poor thing has difficulty crossing the Atlantic (or Pacific.)
"Losingest" isn't another Doctorow coinage but is recorded in a Peanuts strip (Charles M Schulz.)
I don't suppose Schulz imagined the US was to have the losingest President and Administration… probably ever… but here we are.
