At my current academic institution, I have observed an interesting new fracture among physicists. As you may expect, theoretical physics, like many other fields in the United States, is not immune to the growing pressure to adopt artificial intelligence, and my own view of it has shifted along that same fracture line over the past few months. In the first couple of months of 2026, I started using it more after I was first given free, almost unlimited access to it. At the time, I felt fairly good about it: it solved some of my technical problems faster than I would have on my own, and it gave me something close to a sense of progress. Even then, I was roughly aware of the pressure this puts on land (and everything living on it…), energy, and fresh water. Still, I carried a naive hope that the progress I felt (and the hypothetical scientific progress in other fields like medicine) might, in the end, be worth what it cost.

Perhaps AI would also be the first existential threat to cure itself?

As the months passed, my initial hope began to fade, giving way to a deeper, uncomfortable feeling: a growing sense of disconnect and disinterest in the part of my work that involved AI. Ironically, I believe this loss of purpose in practicing AI-assisted physics stems from a lack of questions about the practice, its unclear direction, and its potentially catastrophic effects on the field and society in the long term. (In case it is not clear to the reader, I’m here, of course, referring to the need for discussions among ourselves, humans, rather than simply sending prompts to chatbots.)

A world not worth saving?

While I will come back to AI, I think the main question we’ve overlooked in the last 30 years or so (basically since the United Nations Climate Change Conference was created), and that has led us where we are now, is not really about which technologies to develop or not and not even about how to make our world more sustainable. It’s whether we genuinely want to continue on this particular path at all, sustainable or not. I find this especially troubling that we keep asking how to extend this situation for a little longer (and probably not for much longer, realistically) without first considering whether the path we’re on has ever been worth pursuing.

We tend to reason as if it were obvious that raising children on short videos and feeding them supermarket food coming from the other side of the planet, within a world of concrete, algorithms, and machines that quietly allow refugees to die at closed borders, is inherently the right thing to do.

Then, we focus solely on finding ways to maintain this lifestyle for a few more decades, with slightly reduced carbon emissions, because we have developed a conscience about it. I do not think this is a serious response… We have become so dependent that we can no longer perceive the harmful aspects of our situation, independent of its ecological footprint.

I think we are building a world that would be repugnant even if it turned out to be durable. Things like AI costs energy, certainly, but I think that was never really the point. Suppose the carbon and water accounting came out at zero for some magical reason: would we still be glad to know that a handful of companies hold something close to a complete record of our lives? We could, in principle, win on every axis of life and still not want this. That is the sense in which I do not think giving it up is a sacrifice. Living without it is not living worse!

What happens when rockets become more beautiful than the stars?

Let’s consider the vision of the future being shaped by the few wealthiest individuals on the planet (who are also the ones implementing AI in every aspect of our lives and polluting more, by themselves, than entire continents like Africa.).

They are the outright winners of the global game of Monopoly, creating a universe that is vulgar, violent, and devoid of stars. The main issue here is not just about emissions, it is that the very essence of this future is repellent to me. We see them talk about killing death itself, about brain implants designed to connect our minds directly to machines, social media platforms operating without any real oversight, flamethrowers available as consumer products, and workforces treated as disposable… Even the visible sky I was used to looking at in my youth is now being polluted by satellites. This entire set of choices, and not really their negative effects, is what I believe should deeply concern us and is most likely what is at the heart of the loss of sense and purpose I was eluding to in the intro.

In a troubling final act, we have allowed even the Cosmos to be contaminated. Private launch companies are advertising in orbit and creating fuel depots and hotels circling the Earth. It seems to me this is turning space into a form of indecent tourism for those with indecent fortunes… Constellations are being quietly redrawn by masses of satellites. While conditions on Earth deteriorate, our imagination has shifted toward a kind of planetary exploitation ranging from the depths of the oceans to planetary orbits, a misguided approach regardless of its carbon impact.

I guess this is what happens when rockets become more beautiful than the stars.

Confusing the medication with the poison?

All of this is unfolding at the moment alongside the spread of AI, and here I want to be precise about the question I think we should focus on. It is not about whether to decarbonize AI. At best, that would only recover the loss on the modest progress on emissions we were making before AI came along (it actually apears from a recent study published in Nature that AI-driven productivity gains end up enabling more CO2 emissions than they avoid in a global energy–economy model — what a great example of Jevons paradox at work). It is whether we want systems that to generate a version of our reality that is predictable, standardized, pre-calculated: without exuberance, without surprise, without the chance of anything unplanned, built for surveillance and killing drones and tuned to narrow many societal margins.

Do we want online advertising sharpened to further deepen our compulsions to buy and hate? Content selected for us, by autonomous machines, to confirm what we already believe rather than test or challenge it? Should our résumés be read by a program before a person sees them? A global financial system, which already governs a great deal of our lives, increasingly running on code that essentially no one, including the owners, understands (this is also becoming true in scientific programming, unfortunately…). Our desires anticipated closely enough that nothing outside them can occur? Insurance and risk priced from probabilities we no longer have any say over? With no regard for what all of this actually costs downstream in ecosystems and in the labour of people far away who never asked for any of it? Put together, that list starts to look less like a set of conveniences and more like a single, LARGE mistake.

And whenever the legitimacy of any of this gets questioned, it tends to be filed by a fair number of specialists (yes, scientists are a huge part of it) and decision-makers under the usual umbrella of “obscurantism.” I must say that I am a bit sick of this argument. No one should get to claim that artificial intelligence is a solution to the problems it is causing. It is the problem. It will produce some genuine benefits here and there, sure, the way most carcinogens do.

We should not confuse the medication with the poison.

The usual response, that medicine stands to benefit, is, I think, the “Godwin’s law of Nazi analogies” of this whole discussion. A handful of marginal medical gains cannot justify a technology that is broadly destructive. It is a bit like clear-cutting every forest on earth in search of new molecular compounds. We would, no doubt, learn something useful for medicine and save a few lives, and five years later (this time scale estimate is probably too optimistic), we would all be dead from generalized asphyxia.

To me, all this does not add up and, deep down, I think it is what is causing the loss of purpose and direction I mentioned earlier.

Words without an author…

A chatbot is, in fine, a (fairly simple?) program for completing sentences, trained statistically on whatever text it was fed, tuned to guess the most likely continuation in a passable (questionable, but ok) imitation of how people write. I think the fact that we find this thrilling says something about how little we now expect of ourselves, our lives and the world around us. This, at a moment when the actually marvellous, the actually alive, is dying around us at a considerable pace, is the real tragedy.

One can ask another interesting question, for which some claim (I think, too prematurely) to have a positive answer to. It is whether or not AI can do more than what it actually is by nature: a system built to interpolate. To me, it is not entirely clear that anything genuinely new can be produced (including recent advances in mathematics; I am not sure that all credit for the existing obscure literature on which the proofs and counterexamples are based has been disclosed honestly so far). So, until it does something conceptually groundbreaking, my take on it will be conservative, and it will at most interpolate between what is already known and what has already been produced.

I might be wrong on this last point, of course, but that would not really matter. 

Who should we accuse for the potential harm done by AI? Who should be responsible when no one assumes the words that lead to something catastrophic?

Would we be happy to take the risk of being the generation that exists just to disappear (literally or not)?

Let me end with a couple of questions I would like the reader to think about.

Is democracy itself, along with our rights to accurate information, fulfilling work, creativity, breathable air, and drinkable water, at stake in a world where we let machines creep into every sphere of our lives, think for us and consume some of the finite yet essential natural resources we ourselves need to live?

Are we truly willing to let the same individuals creating this technology and claiming, at the same time, that they lost control over it to the point it is leading us toward an immediate catastrophe, answer this question for us? In other words, are we really willing to lose our ability to create a desirable future?

In many ways, I believe this could be the most significant question we will face in the coming decades. So, why not start thinking about it seriously now?