A recent paper extolling the virtues of AI for deliberative democracy whisked me right back to the early days of the internet when that new technology’s miraculous properties gave rise to countless utopian visions of a more equal world with enhanced citizen participation.
We all know how that turned out; our new interconnectedness and access to troves of information did not usher in a golden age of citizenship.
On the contrary, it is heavily implicated in facilitating disinformation, polarisation, atomisation, centralised control, capital concentration and a slew of other challenges to stable and responsible governance – apart from the undeniable advances it has also brought.
AI is quickly shaping up to exacerbate these challenges, rather than solve them, as some who still believe in a more one-dimensional model of progress optimistically might posit.
Still, this is where there could be room for a more positive engagement with the technology, through regulation and strong government-led and international initiatives to curb abuses, bias, inequalities and the concentration of resources in the hands of a few unaccountable mega-corporations.
Regulatory divide
The clearest example of this is the EU’s AI Act that seeks to address multiple challenges, including deep fakes, manipulated content and AI ‘astroturfing’, i.e. the use of AI to fabricate fake grassroots support for a product, cause, or political narrative.
While mostly aimed at some high-risk AI uses that pose a threat to human health, safety, or fundamental rights, it has also introduced a new category of systemic risk that covers all general-purpose AI models from large corporations.
The EU, in its recent Technological Sovereignty initiative, also addresses the issue of the infrastructure needed for AI, mainly data centres, being dominated by a few major players.
The bloc aims to expand European compute capacity to at least 30 per cent of its own needs by 2030 and, more drastically over the next decade, to full self-sufficiency. By 2030 it should also supply at least 20 per cent of its own semiconductors.
The US and the big-tech companies are putting pressure on the EU to water down its AI Act
The logic here is not just that the EU cannot be dependent on the US or China for what is seen as a crucial new technology. But also, that it can actually gain an advantage by growing its own AI industry on a more responsible, human-led basis.
This is contrasted by the more AI-forward approach pushed by the big-tech corporations and others, including some politicians who see the technology both as a panacea for a range of inconvenient problems and as too important to risk falling behind geopolitical rivals if regulation slows down development.
The latter is exemplified by the US approach, where President Trump recently signed a watered-down executive order asking companies to submit significant new models for a voluntary 30-day vetting period before general release.
The US and the big-tech companies are putting pressure on the EU to water down its AI act, which the Commission did to a certain degree recently, although the main act remains mostly intact.
The Trump administration’s laissez-faire attitude is not shared by all in the US. Even a staunchly pro-business Republican like Florida’s governor Ron DeSantis has thrown his weight behind his state suing OpenAI and its CEO over a “litany of harms” it has allegedly caused, particularly to children.
Developing at breakneck speed
AI chatbots, starting with OpenAI’s ChatGPT, have been with us for merely four years now. The technology is developing at breakneck speed and affecting lives and businesses exponentially faster than the internet did initially.
Unlike the optimistic early-internet discussions, the discourse on AI has become hugely polarised very quickly. Some boosters promise outlandish progress, while elsewhere it has turned negative and angst-ridden.
Fears range from it taking our jobs to our very humanity, even sparking the current pope’s first encyclical.
Somewhat remarkably, those using it the most, Gen Z, are also the most critical, a recent Financial Times survey found.
AI infrastructure is expected to be worth some $3.5 trillion next year
AI evangelism, as propounded by top-tier tech-industry moguls, crypto-bros, many others in finance and, as noted, quite a few politicians, has been met with jeers at commencement speeches in the US, the newspaper reported.
Recent relative setbacks to its rollout are unlikely to stop the technology’s momentum. Uber, for example, has said its outlay on AI is not delivering commensurate returns. And Klarna is hiring back humans in customer service.
For now, though, this looks more like a problem of implementation and fine-tuning than a turning point that will slow down the AI juggernaut.
There is an unprecedentedly vast amount of money being invested in AI. Its infrastructure is expected to be worth some $3.5 trillion next year, of which more than three-quarters in the US. It is now thought to underpin some 30 per cent of the S&P 500 stock valuation, despite some recent losses.
Massive infrastructure investment
This also stands in sharp contrast to the early years of the internet, a system developed mostly by the defence establishment in conjunction with academics and in which organisations such as CERN played a key role in refining the world wide web.
The internet was only gradually taken over by vast commercial interests to make it how we experience it today, both the good and the very bad.
Mostly this has taught us that we become the product, even to the extent that AI has used our own online output to replace at least some of what we used to do.
AI is now starting out from that hyper-commercialised position, and despite some ‘slow AI’, local AI and open-source initiatives is almost exclusively a corporate affair.
The EU is eager to catch up in compute but the bloc’s environmental rules and energy costs make that a thorny task
China’s DeepSeek briefly appeared to disrupt that paradigm, with reports of it having been developed ‘on the cheap’. But this was undermined by its ties with the Chinese government and reports of it building on, or even copying from, Western models.
The absolutely massive infrastructure investment needed to run AI, mostly data centres and electrical grids, forms a huge barrier to small-scale initiatives being able to compete.
The data centre energy conundrum also feeds into a whole range of other environmental and geopolitical issues.
The EU, for example, is eager to catch up in compute but the bloc’s environmental rules and energy costs make that a thorny task.
Limiting our options rather than expanding them
Having a high-impact new technology being so dominated by relatively few ultra-wealthy individuals and companies is contributing to the widely felt sense of dissatisfaction and loss of control that many people, particularly in the West, now experience.
This goes doubly for those just starting out in the job market or currently in school, who rapidly see traditional career paths narrow and also face other pressures such as the cost of living and housing.
Attempts at regulation by the EU and others already lag behind and risk being inadequate
AI adds insult to injury, presenting most of us with a faceless, corporate or bureaucratic excuse for limiting our options rather than expanding them – and which too often offers no human recourse.
This makes the question of whether the EU’s model – of a responsible and equality-based approach to AI – is realistic all the more important.
We can by now be quite certain that large tech corporations will not follow adequate safeguards if not forced to do so, either by regulation or a threat to their profits, preferably both.
This is becoming even more urgent in light of the latest AI developments, with several labs reporting they are on the verge of creating models that can improve themselves. Concerned safety experts warn this could produce ‘Terminator’-like machines.
Whether this is alarmist or not, several things are very clear: The tech industry itself will not hold back, nor will it install the necessary safeguards.
Attempts at regulation by the EU and others already lag behind and risk being inadequate.
And lastly, also judging by how big-tech won the fight for the internet, small-scale, independent initiatives to ameliorate AI’s effects are unlikely to succeed.
The real issue around AI is not to how to invent ever more fanciful utopian uses, such as it enabling deliberative democracy. The question is how to steer it away from a plethora of ever more plausible dystopian futures.
