__ __ _ _____ _ _ _
| \/ | ___| |_ __ _| ___(_) | |_ ___ _ __
| |\/| |/ _ \ __/ _` | |_ | | | __/ _ \ '__|
| | | | __/ || (_| | _| | | | || __/ |
|_| |_|\___|\__\__,_|_| |_|_|\__\___|_|
community weblog
AI is by its nature an imperial technology
The great corporation-states of the European imperial era—the Dutch and British East India Companies—minted their own coins, fielded private armies, levied taxes, and bargained with kings as equals. They drove imperial domination by creating trade monopolies, defending them with military force, and ruthlessly exploiting the periphery. Today's counterparts are frontier AI companies such as OpenAI and Anthropic, which command the human talent, data, models, and finance on which advanced intelligence depends. They are not merely aggregators of capital but arbiters of knowledge, increasingly determining what is discovered, how quickly, and to what end. [Foreign Affairs; archive]
posted by chavenet on Oct 25, 2026 at 4:25 AM
---------------------------
The names on it that I recognise are people I don't trust (Niall Ferguson's reading of empires is in the neoliberal school and Eric Schmidt gave up "don't be evil"), plus it has a citation of Nick Bostrom ... but I think it's fairly well-written and covers a broad section of the topic.
I think the predictions of "dead internet" as a failure mode are a bit nuanced for this piece, but the influence ops of advertising (used for political ends) or algorithmic endless scroll are other avenues for hegemonic shaping that AI supposedly accelerates. The most engaged, most valuable customers are often the same people who've taken the most deliberate steps to make themselves less measurable by conventional ad tech. So those best consumers might hide under "invert the killfile" response to ad-tech amplified by AI, at the cost of eroding the hegemonic power of these AI empires.
posted by k3ninho at 12:14 PM
---------------------------
I could not disagree with this thesis more.
The idea that America and China have some lasting imperial power around AI is risible. AI is about as an open a space as anything this important can be. The science and technology of training, inference and distillation are public domain. Moore's Law doesn't permanently fail to apply to GPUs. There is literally no industry in the history of mankind where capital is or has been more easily available.
India, Israel, Taiwan, Japan, Australia, South Korea, Germany, and the UK can have within-shooting-distance of frontier caliber AI as soon as they need or want it. Ditto Russia if and when it settles a peace with Ukraine and gets its house in order. Countries with more money than know-how (Saudi Arabia, UAE, etc.) might struggle but they get there if they have to do. Poorer and less technologically sophisticated countries are going to have to license but it's not like licenses won't be available.
posted by MattD at 12:27 PM
---------------------------
I'll have to dig into this. I'm reflexively worried about agreeing with aything written by Niall Ferguson. But there's a core.piece here that AI exists to replicate the structures of power, and for all the boosters claims of its so-called freedom, it's main impacts and desires are to elevate capital over labor. I wonder if it also is now something unifying the class consciousness of those with money above even national interests. The way we've seen these bad technologies pushed and adopted makes it clear though that the drive behind them is inherently antihuman. In a large way that's actually why it's such an imperial technology.
posted by Carillon at 12:46 PM
---------------------------
I think 'neofeudal' rather than 'imperial', Carillion, but that's possibly splitting hairs.
I can't recreate, improve or validate the local models I've downloaded, and I'm at the mercy of the people (stealing the zeitgeist then) distilling and reworking it into the online-only GenAI tools. If these become the predominant ways to function in the workplace, I'll need to pay company scrip to go to work, and further misery from rent-seekers will follow.
posted by k3ninho at 3:39 PM
---------------------------
I distrust Eric Schmidt because of the enshitification of Google, as well as his efforts against Assange, but maybe he'll say something interesting sometimes. Niall Ferguson is an apologist for the British Empire, but at least he has real knowledge about empires I guess.
"Lisa Klaassen is a DPhil student at the Oxford Internet Institute, specialising in the nexus of artificial intelligence and national security. Her research examines how technology companies shape the discourse and governance of AI within the context of US-China strategic competition, and explores the implications for middle powers. " Awfully suspect background, because way too many people talk national security without understanding anything about harm reduction component. And her lawfaremedia.org post says relatively little.
Do these people realise that LLMs help developers find and patch exploitable bugs in exactly the same way that the help attackers find exploitable bugs?
There is an inherent security risk caused by not patching bugs. Yes, LLMs speed up the mean time to exploit, but they speed up fixing bugs and enable many other tools, assuming you bother spending enough money fixing bugs.
Also Nick Bostrom is a deluded fool. I doubt her understands that security nuance. Actual information security is not a part of his little fairy-tail land.
posted by jeffburdges at 4:05 PM
---------------------------
AI guardrails aren't stopping criminal hackers, but they are stopping developers from detecting and preventing exploits.
posted by jeffburdges at 6:56 PM
---------------------------
"Niall Ferguson's reading of empires is in the neoliberal school"
In that light, this is urging the US to abandon social welfare programs in order to save AI and preserve hegemony.
posted by blue shadows at 10:55 PM
---------------------------
Niall Ferguson is a proof an idiot with an education is still an idiot and it's fine to automatically disagree with anything he says.
posted by aspersioncast at 6:02 PM
---------------------------
I'd prefer to call them LLMs as a variant of general AI tech, but I think I've lost that battle already. It's just delusional to say they have no serious application at this point.
posted by chmmr at 6:28 AM
---------------------------
I avoid the term AI and call them LLMs for several major reasons, including that future AI architectures shall beat LLMs.
In particular, architecturally LLMs are nearly deterministic pure functions from their context window, query, etc, over which the operator has control, so they cannot "learn" anything except by retraining, which often costs billions.
LLMs wind up non-deterministic only because of the rounding in floating point arithmetic combined with threads finishing at different times (see also 2408.04667), but could be made deterministic, like by using integer arithmetic.
posted by jeffburdges at 3:58 PM
---------------------------