Europe 2031: A Response, and Call for Open-Ended Inquiry
Tl;dr
This is a response to Europe 2031, a scenario published recently with significant resonance across the European policy ecosystem.
Europe 2031 is a scenario-driven argument: In order to avoid being powerless vis-a-vis US and Chinese geopolitical and geoeconomic coercion, Europe should heavily invest in data centre construction on European territory. Having data centres to run advanced AI models preserves European autonomy in an age where AI drives productivity and economic competitiveness. (more below, or just head to the authors’ own summary[1])
My response complains that this argument rests on assumptions[2] which are left implicit in the narrative, when instead they should be made transparent. There is much that we do not understand well about how Europe can respond to accelerating AI development and power concentration, and we need open discussion about those uncertainties and competing responses.
I respect the authors of Europe 2031 for their genuine effort to wake up European policymakers to a forward-looking response to advanced AI development. But I also think that publications like Europe 2031 are net negative in their effect on the world if they foreclose discussion rather than encouraging it. This post is an attempt to push further in the encouraging-discussion direction. It’s meant as a complementary effort to what the scenario is trying to do, not an adversarial one.
In that spirit, here is the overarching question I would love to see discussed more[3], rather than less, amongst all those who engage with Europe 2031’s call for more serious EU AI policy endeavours:
How do we overcome the coordination problem of global AI governance? How do we prevent extreme power concentration within a few leading companies (and individual company CEOs)? What role can Europe play in responding to these challenges?
Also in that spirit: I’m listing a few resources that have started these discussions in this GoogleSheet: EU_AI_geopolitics_resources. If you know of or encounter others, please do drop them in the comments or submit them via this form, and I’ll endeavour to keep the sheet a living resource repository going forwards.
Context and epistemic status
This is a brief essay about my concerns regarding Europe 2031 and the way it may be read and received. I was a bit reluctant to write this because I understand and share the authors’ frustration with European inaction, which is in part fueled by internal division and endless bickering. But I’m also reluctant to not write down what worries me: the publication of Europe 2031 may have net negative consequences if we don’t receive it critically and if we use it as an end rather than a start to discussions around the EU’s geopolitical AI strategy. I hope that this essay helps us think more critically and intentionally about the futures we envision and about how we work towards them. It is very tentative, not presenting final solutions but raising questions and prompting us to look more carefully before locking in onto any one strategy. I mean for this to be a critical but constructive contribution to the effort Europe 2031 has started, not an adversarial attack.
I spent roughly 3-4h drafting, discussing with friends, and formatting the essay below. It could be much better, more thoughtful, and more incisive in the arguments it is making (especially in its identification of assumptions in Europe 2031 and its outline of possible alternative perspectives). I am keen for criticism, refinements, and expansions of the ideas I start to develop below; as I emphasise above: the goal is engaged discussion, including discussion that takes apart what I’m saying!
The actual post
Intro and main argument
Before launching into my critique, let me note what I appreciate the scenario for: it points at what I feel are valid and needed concerns about EU officials and institutions being indifferent to AI and its possible geopolitical implications, about the incredible lethargy of European bureaucracy, and about people with authority unjustly dismissing thoughtful input from lower-ranked colleagues. I share a sense of frustration with all of these, as well as a desire for things to improve. And I think Europe 2031 may be an effective prompt to bring some such improvements about, which I feel grateful for.
Why, then, do I worry about net-negative outcomes from the publication of the scenario? Because I think that it unreasonably presents one kind of AI risk and one bucket of solutions to mitigate that risk, making the implicit case that this is the one risk we need to be most concerned about and hence the one bucket of solutions we need to push for. That implicit case is—or should be—deeply controversial, based not on solid evidence but on a bunch of value judgement and empirical hunches. Making such value judgements and empirical guesstimates is fine; inevitable, even, if one is to venture into the space of policy recommendations at all. But presenting them hidden underneath a well-written and captivating story, rather than making the case openly and explicitly and with epistemic humility and transparency where warranted, is not inevitable.
My reading of what Europe 2031 is telling us
The one risk that is presented in the scenario runs roughly as follows: frontier AI is presently being developed mainly by a few US companies, with Chinese competitors reasonably close behind whilst European-developed models remain far from the frontier. That makes European end users dependent on access to foreign AI models. As AI integration starts to dominate growth and productivity potentials around the world, dependence on foreign AI models makes Europe a vassal of whoever it can rent these models from. Such dependence basically disenfranchises Europe as a sovereign entity because the providers of frontier AI models can derive crushing leverage from Europe’s need to deploy frontier models. A disenfranchised Europe is a Europe without a viable future, and certainly one that is incapable of defending its values and European citizens’ interests in the decades to come.
The bucket of solutions, if I read the scenario correctly, mainly revolves around the building of compute capacity on European territory: data centres primarily, possibly also a greater portion of the supply chains needed to build the semiconductors that go into those data centres, with the goal of providing model access and model development capacities through European infrastructure.
A critique of what Europe 2031 is telling us (and of how it tells us)
Europe 2031 tells a story that is not without merit, nor one built on demonstrably false premises. But it is a bit arbitrary: other stories would be possible that would seem equally meritorious and would equally rest on not-demonstrably-false premises. Alternative stories could, for instance, depict a dystopia where a massive European data centre build-out sacrifices regional standards and distracts from global governance efforts to slow down AI development, while at the same time doing little to reign in the power of the United States (which continues to house leading frontier AI companies) over Europe. Or, to pick one other among many possibilities, such a story could portray a future where access to the newest AI models is only marginally important because progress in frontier models easily leaks out to help improve replica models (similar to what we seem to be seeing with Chinese model development at the moment), which means that industrial productivity, innovation, and research capabilities are nowhere near as badly affected by US frontier model export controls as Europe 2031 insinuates.
The choice of which of these stories to tell is arbitrary rather than random or coincidental because there is a specific worldview that underlies the narrative adopted by the authors of Europe 2031: one that privileges deregulation’s short-term gains for innovation over its longer-term risks to societal and environmental sustainability; one that accepts competition as an inevitability and holds that collective action problems, at least at a global scale, simply cannot be overcome, no matter the costs of failing to overcome them; and one that has accepted the risks of rapidly evolving AI development and thus absolves even self-avowedly responsible actors from the burden of fighting such recklessly fast development.
Building more data centres with fewer regulatory hurdles in the EU is not the one sensible response to our emerging AI-shaped geopolitics. It is a heavily assumption-laden response: It assumes that more data centres really will make the EU more autonomous and immune to US coercion, even though leading AI companies will continue to sit in the US (thus giving the US leverage even if it does not have a near-monopoly on compute). It assumes that more data centres in the EU will not speed up capability development, or that sped-up capability development is not a risk worth fighting (despite dire scenarios about the possible ramifications of AI models that improve faster than society can adapt and faster than safety researchers can build guardrails). It assumes that material assets are the only real instrument of power, the only truly secure option to protect the EU’s sovereignty and global influence. And it assumes that the conflict exists between countries (or country blocks), rather than between power-amassing monopoly companies and everybody else.
I can’t help but feel that following the recommendations laid out by the authors of Europe 2031 is giving away much of what we should be fighting for: a chance to mitigate extreme risks of AI; a chance of preventing extreme power concentration in the hands of a few recklessly accelerationist companies; and a chance to think critically about pathways to AI development where technical progress does not outpace the wisdom and resilience we need for a future where technology actually serves flourishing, rather than dooming it.
A plea for rigorous and honest discussion
In other words, the questions I’d love to see contested and discussed, in cooperative but non-groupthink-y ways:
What exactly is the pathway through which increased computing capacity ensures European sovereignty? What are the assumptions required for that pathway to work, and how uncertain are those assumptions?
Does increased demand for data centre production influence the speed of AI capability improvements?
What are alternative strategies for advancing European values and interests in an age of advanced AI development?
How can the EU play a role in reducing the catastrophic risks from unsafe model development?
How can the EU reduce the systemic risks of AI integration in its economy and society?
What level/type of regulation and public action is required to implement these strategies?
More broadly: How do we overcome the coordination problem of global AI governance? How do we prevent extreme power concentration with a few leading companies (and individual company CEOs)? What role can Europe play in responding to these challenges?
A living repository, and continued discussion?
In the spirit of wanting to spark and advance discussion, I’m collecting a few resources that speak to the questions and arguments above: EU_AI_geopolitics_resources. If you know of or encounter others, please do drop them in the comments or submit them via this form: https://forms.gle/1tW1VAE9tA9QrxSKA. I’ll endeavour to keep this a living resource repository going forwards.
***
Lastly, as a PhD researcher studying the epistemics of foreign policy in a context of inconclusive evidence, I’m always keen to meet and convene policy analysts working on some of the most tricky knowledge problems we face in trying to tackle global challenges. If you are one of those, and might be interested in discussion exercises or joint projects with other researchers trying to make sense of Europe’s role in global AI governance, do reach out: sarah.weiler@strath.ac.uk.
- ^
For a summary of the scenario itself, focus on the first few sections. The last section of the linked summary outlines a range of policy recommendations, some of which go beyond what is discussed in the scenario itself (as far as I recall/understood). That last subsection fits neatly with my call for transparent argumentation and broad discussion. My argument in this essay is that this transparency and honest argumentation is lost in the scenario itself.
- ^
This post is a quick write-up, so no claim to exhaustiveness, but I identify some of the assumptions that strike me as particularly contentious and debate-worthy below.
- ^
A longer list of (sub-)questions in the concluding section below.
Hi Sarah! I have more to say on this, and this comment may sound more dismissive than I intend it to, because I don’t have the time right now to respond in full (and I’ll be honest: I can’t promise I’ll make the time later either).
But a few things I’d say for now:
1. The scenario is a story, and all narrative scenarios are. It is not presented as inevitable, otherwise there would be no point in writing it. It’s a possible world we could end up in if nothing changes in Europe. The point is not to paint a full analytical picture; the point is to change minds and inspire action. So I don’t think it’s quite right to hold it to the standard of a rigorous analytical exercise: it’s a different tool for a different job. The fire alarm is not the same as the fire-safety report. Of course the questions you raise are absolutely legitimate and important to discuss, but I would put back to you: have you actually seen less of this kind of discussion since Europe 2031 came out? Or has it sparked more debate about how to generate European agency, and what we should use that agency for?
2. The scenario clearly describes itself as having a narrow focus on purpose. The scenario is exclusively about maintaining European agency, and says nothing about how to use that agency once it has been achieved or strengthened. But that is a scope choice, not a statement of what the authors care about. The FAQ addresses your concern almost directly:
3. On power concentration and international coordination: These are super important, but you can’t do much about them without agency. Europe being “AI colonized” and losing its agency to a handful of foreign companies and states, is itself one of the forms of power concentration the scenario is trying to get Europe to defend against. Building European leverage is, in part, a hedge against exactly that concentration, as well as a way to gain the agency to combat other problems.
For a nice description of how unreasonably effective this kind of fictional account can be, I’d recommend this post too.
Hey there!
Thanks for responding, and for your honesty regarding time for engagement. If you end up writing out in greater length what you have to say, happy to see an extended version of your response. That said, it seems to me that what you did write already provides good grounds for me to engage, so here’s a
quickresponse from my side point-by-point:I think I largely agree to this, actually. That’s why I tried to emphasise in my intro, and to a lesser extent throughout the post, that I see my response as complementary: narratives are a good tool to inspire action, but they require attendant analytical exercises to have positive value. Action in an ill-considered direction is worse, I would argue, than no action at all, and definitely worse than reflective action.
The scenario Europe 2031 (in contrast to other narrative projects that make more of an effort to highlight uncertainties and multiple possible futures, see e.g. AI 2040) leaves its readers with a fairly strong conclusion in the Epilogue, one that very much sends the (emotionally charged) message “we know what we need to do, and we should do it even if it is politically painful”. The authors offer a more analytical, and somewhat more nuanced / uncertainty-aware, outline of their argument in the Summary, the FAQs that you also (rightly in my mind) cite, and their Compute Forecasting section. But my sense is that many readers will mainly focus on the scenario itself, not the analytical bits on the side; this post here is one attempt to shift that a little bit further in the direction of using the scenario as an entry-point to the analytical discussion, rather than as a shortcut to a seemingly clear conclusion.
You are right, I think, to ask about the actual empirical effects of Europe 2031′s publication. The honest answer is (a) that I’m not in Brussels (or another EU hub), so I have a very poor sense of what is being discussed with how much of a truth-seeking intention to it; and (b) I think it’s too early to tell. I do think, however, that other people have established quite well how narratives and stories can obfuscate public and policy debates[1]; and that feels enough to at least be worried about Europe 2031 having a similar effect, if we don’t try explicitly to counter that obfuscating effect.
That’s largely true, but I also think it’s a good example of means and ends being hard to separate in practice. “Maintaining European agency” is a means to an end which the authors barely discuss, and I think it’s fair for them not to discuss it at length (for the reasons they themselves list in the FAQ, and which you cite above). However, to the extent that their proposal for “maintaining European agency” hinges on instrument that have implications for diverse AI governance aims, I do think they have a responsibility to discuss that; or, at least, I think we should use the scenario as a starting point for discussing these implications. If the scenario’s proposal for “maintaining European agency” mainly consists in “massive compute build-outs on EU territory” (and I have confirmed with one of the authors that my reading of that is broadly in line with what they themselves take the main message of the scenario to be), then that is not neutral with respects to “other [deliberately ignored] aspects of the AI transition”: such a build-out might accelerate frontier AI development; might further amass power with US frontier AI companies (if, as the scenario suggests, such a build-out would happen under the auspices of these capital-rich US frontier companies); and might undermine EU regulatory standards, which could otherwise be used as a tool in AI governance efforts. It also diverts capital and attention away from other potential European responses to rapid AI developments.
I think my main sticking point here is: I agree that “maintaining European agency” is a worthwhile instrumental goal to strive for, because I too would like to the see the EU as a capable actor in future AI governance developments; but I think that we face many really difficult and politically/normatively meaningful questions when it comes to deciding which strategy to pursue in order to maintain European agency. I want our response to Europe 2031 to be that we discuss vigorously and honestly what different strategies might look like, and what their merits and costs are. I think the scenario itself may not spark these discussions (because neither the scenario nor the authors’ supplementary commentary insinuate that there are open questions here at all). So I wrote this Forum post to start filling that gap.
My response to this is largely the same as the one offered above: I think there should be controversy and open discussion around how we can prevent Europe from being “AI colonized”, around what it means for Europe to build up effective leverage, and around what we are and are not willing to give up in our attempt to build up such (supposed/potential) leverage. My sense is that Europe 2031 “falls into the same trap”[2] as Anthropic itself did: assuming that a race is inevitable and that the best option is to come out on top (or near the top, in Europe’s case), even if that means accepting all of the costs the race brings in the first place. I want for that assumption to be vigorously debated, not taken for granted.
Lastly, thanks for linking the post on the (potential) effectiveness of fictional accounts. I haven’t read it yet, but it definitely falls within what I’m interested in and I’m glad to have the link (and also for others visiting this post to have access to the link)!
Here I would insert a reference from the study of propaganda, media studies, or a similar field if I had the time to look for it. I hope to come back to this later this week, but apologise in advance if I fail to. My only excuse is that those fields and their general takeaway—stories influence minds, and not always in an epistemically ideal way—is a sufficiently well-accepted piece of common knowledge that my claim survives without a specific reference (?).
In quotation marks, because it is only a trap given the specific worldview from which I am speaking; one that should, just like Europe 2031′s worldview, be vigorously debated and contested!
“where a massive European data centre build-out sacrifices regional standards “
Can you say a bit about what you have in mind here and why it’s important?
Hi Sarah,
Thanks for the write-up. You are more patient than I in trying to engage with this in good faith – my initial reaction upon reading the doc is that a policy proposal on EU tech sovereignty that fails to propose anything the likes of Marc Andreessen and David Sacks would disagree with is not an argument made in good faith. I must say that I’m disappointed in many ways by the overall quality – both of the narrative writing and the political modelling behind it – which is why I’m trying to hold myself back from getting nerd sniped into critiquing every little detail in the document. Instead I’ll focus on two arguments that could most help fruitful discussion:
The tension between the article’s fearmongering about Europe as a US vassal state and their proposed solution of subletting European resources to US AI companies.
The geopolitical argument: is Europe weak or strong as a non-aligned middle power in the bipolar US-China AI buildup?
On the build lots of data centers proposal
Start with the central incoherence: Europe 2023 spends multiple acts dramatizing the horror of vassalage – country-tiered inference, the American AI valve that can be closed overnight, a handful of people in San Francisco and Washington deciding humanity’s future. And then its flagship remedy is to build more American infrastructure on European soil, faster: dedicated economic zones, streamlined permitting, partnering with American providers on a vague promise of guaranteed access to frontier AI. “Under European jurisdiction” is carrying a lot of weight in their proposal, but any MP concerned about tech sovereignty would ask how a data center physically located in Ireland or the Netherlands, running US silicon, executing US foundation models, orchestrated by a US cloud hyperscaler and managed by a US tech giant is a worthwhile step towards a stronger Europe. In fact, I’d argue the opposite: fast-tracking data centers doesn’t disentangle Europe from the American ecosystem, but sets that entanglement in stone for the coming decades. A relevant excerpt on the actual realities of European digital governance from Antoine Vauchez (source below):
And its closing note:
This points towards a morbid inversion of Europe 2031′s proposals – the relative weakness of European tech is currently being exploited by an aggressive US public-private industrial lobby to effect the exact deregulatory agenda that they proposed.
On Europe as a middle power
This is much less my specialty, so I’m framing it as a question for others to engage with. Europe 2031 treats non-alignment in the AI race as pure weakness, but isn’t this backwards from how middle powers usually fare in a bipolar contest? One could argue that Europe’s strength is in its relations with both China and the US, which are better than US-China relations, as well as its wealth and large single-market, making it an attractive customer to AI companies desperate for steady revenue streams.
So why does the article frame Europe’s lack of local frontier AI as complete disempowerment? My sense is that the AI race frame is an assumption underlying the whole deal that does a lot of the work in carrying the argument to fruition. A race has one axis and a linear ranking, so the moment you accept it, a middle power’s multidimensional leverage – market access, human resources, industrial capacity – becomes moot as long as it doesn’t project them to the top of the leaderboard. Watch what happens to ASML in the story: the one bottleneck both rivals depend on shows up in the final act as a liability, the thing Washington moves to seize, rather than as the strongest card on the table. That inversion feels like an artefact of the frame.
From an epistemic standpoint, I think the AI race framing is worth carefully navigating as effective altruists. Europe 2031 evokes a very two-years-ago view of the AI debate – people who believe in exponential improvement and singularity on one side, people who predict an imminent AI bubble crash on the other. The general position of EA AI safety people is more towards powerful AI than the average academic. I’d argue the AI race framing is a second, independent axis from this debate: whether the race resolves to a clear endpoint with an all-powerful winner, or whether “victory” is multidimensional, ambiguous, and spread across several players. This is an axis on which we, as EA’ers, don’t have to take the extreme position by default. You can think AI is revolutionary and still predict the outcome to be a messy, distributed equilibrium rather than a single company in San Francisco flipping a switch. Europe 2031 is firmly convinced of the linear victory condition, but most of the people whose work I read reject this finish-line framing. However, this assumption radically changes the optimal action for Europe to take – staking out a position as a middle power without resorting to vassalization to US tech interests becomes a lot more feasible in a multi-faceted world with several powerful AI operators.
Some additional notes because I couldn’t resist:
My impression, as someone who has read both Silicon Valley diatribes and conventional policy documents, is that this document doesn’t accurately imagine the actual state of European governance, and harkens closer to the stereotypes of Europe perpetuated in viral Twitter threads. From an epistemic standpoint, EA has historically been about combining the good of Silicon Valley epistemics with the good of academic knowledge; Europe 2031 doing the opposite is a worrying trend.
Specifically, the relation between this article and the current centre-right/far-right push for deregulation through the Omnibus directives is unclear – what should be a clear victory directly relevant to their policy goals is barely acknowledged and tsk-tsked away as “too little too late”.
The article shies away from addressing political factions, but this would be an important step to actual political modelling. The deregulatory agenda largely aligns with EPP, Renew and the industrial lobby, but they would not be reliable partners for the delicate balance the article imagines to be needed for the only possible good outcome (keeping firm enough in negotiations to retain European jurisdiction of what’s going on in the data centers, without triggering US pressure campaigns to give up strategic advantages like ASML early). Greens/EFA are hard opposed and no effort is made to approach them. Social democrats / unions are interesting swing players here, as they are strong voices in the current tech sovereignty push, and their reluctance to concede on labour matters could be overcome with promises of economic plenty. Yet, the article’s writers fail to seriously engage with this political perspective.
Recommended reading:
Hannah Ruschemeier (Tech Policy Press) - “AI hype and the capture of EU AI regulation”.
defines hype as a political object and discusses its effects on the regulatory agenda. Highly relevant and much better written than my own post. Excerpts:
- “Democratic consensus-building is often time-consuming; in the context of digital regulation, this is frequently used to explain the ‘legal lag’—the notion that law is slow and perpetually trails behind technical development. Hype works through speeding up processes, urging stakeholders to take hasty decisions. It creates urgency and pressure to act instantly, in contrast with legal thinking that rationalizes through procedural and structural reasoning.”
- [..] “Current political discussions and regulatory proposals in the context of AI reflect the success of what I call the “innovation narrative” around AI: AI, simply put, means innovation; innovation is desirable without restriction; and regulation hinders it. This narrative is fueled by the misleading picture of an “AI race” among the US, China, and the EU, as well as by the epistemic capture of AI.”
- [..] “Big Tech companies—paradoxically both warning of AI’s disruptive potential and driving its advancement—exert substantial influence over the production of regulatory knowledge through scientific publications This economic dominance manifests in a subset of regulatory capture that I call “epistemic capture.”″
- [..] “Europe’s competitive edge lies in leveraging its unique regulatory identity to foster investment in infrastructure and SMEs, particularly within the strategic frontier of industrial AI.”
Antoine Vauchez (Social Europe) - “Omnibus to nowhere: the quiet dismantling of European Governance”.
stakes out a position directly opposite to Europe 2031
Frederike Kaltheuner, Leevi Saari (Tech Policy Press) - “Europe’s AI Sovereignty Problem Runs Far Deeper Than Frontier Access”
direct reaction to the EU 2031 document that at least tries to engage with some of the realities of the EU political playing field
Dave Karpf—“The data center bankshot”.
this one’s more explicitly left-wing than others, but important to note as it shows clear political goals behind data center opposition that ostensibly well-meaning abundance-type analysts have thus far failed to engage with
https://www.socialeurope.eu/omnibus-to-nowhere-the-quiet-dismantling-of-european-governance
What do you think the important claims being made in this piece are, and what evidence do you think it is giving for them?
I’d be happy to help you if you have a more specific question. For me, the most relevant facts-based claim in the article is here:
This shows that Europe 2031′s main proposals for preventing the “AI-induced decline of Europe”, deregulating the AI sector and fast-tracking data center permitting, are being actively lobbied for by the Trump administration and the Big Tech lobby, which makes it unintuitive how Europe 2031 suggests those same measures as a bulwark against American pressure.
The normative claim made further in the article, is that this deregulatory activity is the sign of a weak Europe conceding legislative territory to powerful interests, and that the way towards a stronger Europe is to defend against this pressure and uphold the laws that protect European public values.
This is only part of the point you are making, but worth pulling apart: the Omnibus processes, which have slightly deregulated AI, are a different thing from the fast-tracking and permitting of data centres, which as far as I understand has received no interest from the US government.
I’d also add that the authors want Europe to stay economically competitive through using AI, whereas the US government is looking out for its own companies. And yes, some of the AI Europe adopts will be American-owned. But if that adoption keeps the EU economy strong, the two aren’t necessarily in conflict.
That’s a fair point, I’ve put a lot of focus on the below point out of Europe 2031, but there may be other interpretations
Hi pmanen,
To be fair, I know a few of the authors of the scenario tangentially, and one of them more closely. Easy to engage in good faith when I feel confident that the authors actually do care about the topic, are not bought by Silicon Valley, and are not in any way as reckless as “the likes of Marc Andreessen”. While I agree much more with your takes and policy position than their’s, I do want to emphasise that I think they very much deserve engagement in good faith, and with an eye to exploring strategies in an uncertain world constructively rather than dogmatically.
I haven’t looked in any detail at the modelling they did, so can’t comment on your skepticism with regards to that. On the two more substantive points, there’s not much need for me to comment, because I largely agree that the arguments the authors on these two fronts make are dubious (/underdeveloped). I would be curious to hear how proponents of massive data centre build-outs in Europe (or other “democratic middle powers”) would respond? Can we get to the crux of what divides serious analysts on this topic, and is there any way to make progress on resolving that crux (or, at least, understanding which irresolvable value differences sit at the core of it)?
I appreciate the bullet points at the end that raise the political angle intertwined with the discussions I’m calling for. I didn’t emphasise that angle a lot in my original post, because I hadn’t thought about it a lot; reading your quick notes, I think it does merit thinking about, so that’d be worth adding to my list of “questions I’d love to see contested and discussed”. Are you okay for me to take your notes as input to make an addition in the original post (would add a footnote linking to your comment as a source)?
Thanks for the recommended readings! I gave them a quick skim for now, and have added them to my own reading list as well as the still-very-incomplete-and-provisional living resource repository mentioned in the post: EU_AI_geopolitics_resources.
Good to hear from your perspective, I admit that my views are shaped by my specific academic direction as well.
Feel free to use anything you deem useful!
Some further thinking on the crux of it all: what irks me is that I can concede the general EA case for AI safety (there is a non-zero chance of highly dangerous superintelligence, and this justifies investments in moonshot AI safety ideas), but that the mindset of Europe 2031 hinges on a narrower case, where:
a) AI becomes so advanced that it would devastate economies that don’t have access to the most advanced model (compare with manageable current-day inequalities, e.g. the West having exclusive access to advanced lithography, Denmark having the exclusive right to produce Ozempic)
b) a competing company / country’s AI is not sufficiently close in capabilities to mitigate that devastation
c) the AI does not become Bostrom-style superintelligent (that would introduce a level of abundance that makes this competition model irrelevant), and
d) the probability of all this happening is high enough that it requires drastically reshaping the priorities of Europe’s legislation and economy