A giant leap … in the wrong direction

Paris, January 2025 in Writing

I have been meaning to write on this issue for quite some time now. To let you hear the alarm bells that go off in my head when my thoughts turn to Artificial Intelligence (AI) and its imminent large-scale deployment. The latest technological revolution to have upended our professional and personal lives will soon be setting standards for data and information consumption.

To let you hear the alarm bells that go off in my head when my thoughts turn to Artificial Intelligence (AI) and its imminent large-scale deployment.

There is a lot I could write about the risks of massive reliance on AI. But what really troubles me at the moment is the near-total absence of debate on the voracious appetite for electricity of this technology, whether in its trial phase or every time each one of us uses it now.

But what really troubles me at the moment is the near-total absence of debate on the voracious appetite for electricity of this technology

The « Artificial Intelligence Action Summit » will open its doors in Paris in February 2025. Organised by the Élysée Palace at the instigation of President Emmanuel Macron, this event will focus on five major issues, none of them related to the energy consumption of this technology. Has the AI pipe dream made us forget that, in 2015, the same city hosted COP21, the United Nations climate conference that adopted the Paris Agreement on Climate Change?

According to Bill Gates, AI is such a game changer that its consequences will dwarf those of the invention of the PC, mobile phone and internet. It is tempting to think of ChatGPT, the most visible tip of the AI iceberg today, as an especially powerful type of search engine capable of handling any question. I recently experienced this, after telling a group of people about a rather poetic – and I hope premonitory – dream. Air France planes were now non-carbon emitting airships and I glided silently across the skies, comfortably ensconced in a well-staffed gondola powered by air currents. I had barely finished telling my dream to my guests when, to my astonishment, one of them showed me an image illustrating the scene I had just described. Her first impulse, on hearing my vision, had been to ask ChatGPT to picture it for her.

According to Bill Gates, AI is such a game changer that its consequences will dwarf those of the invention of the PC, mobile phone and internet.

According to the International Energy Agency, a typical ChatGPT request consumes ten times more energy than a Google search. To put this even more starkly, a study by the University of Copenhagen estimates that the electricity consumption of every single ChatGPT request would be enough to charge forty mobile phones. The fact that it is not possible to verify these figures – OpenAI does not make this type of information available to the public – does not make them less of a concern.

The electricity consumption of every single ChatGPT request would be enough to charge forty mobile phones.

In 2020, Google publicised its objective to reduce its carbon footprint and reach net zero by 2030. Its annual CSR reports seemed to chart its progress until Google’s holding company, Alphabet, put us right last summer, dashing any last glimmer of hope we might still harbour: in reality Google’s carbon emissions had gone up by 48% since its baseline year of 2019, and by 13% in 2023 alone. This spectacular reversal can be blamed on the energy cost of training and deploying AI. Alphabet now claims that it was overambitious of Google to aim for carbon neutrality by 2030.

Alphabet now claims that it was overambitious of Google to aim for carbon neutrality by 2030.

In 2024, Microsoft also revised its ambition to achieve “net-zero emissions” by 2030. Indeed, figures show that its carbon emissions rose by 30% since 2020, in the wake of its partnership with OpenAI. Both examples suggest that the generalisation of AI will put an end to the digital sector’s carbon neutral targets. And the more the energy consumption of that sector soars in response to AI, the more difficult it becomes for the rest of us to tackle climate change.

The generalisation of AI will put an end to the digital sector’s carbon neutral targets.

Did you know, for instance, that Microsoft’s insatiable appetite for energy just drove it to enter into an agreement with Constellation, a massive American energy corporation, to reopen the notorious Three Mile Island nuclear plant?

GAFAM are racing to lay their hands on carbon-free nuclear energy. Yet the AI systems they are so aggressively rushing onto the market will be old news and the Paris climate agreement in tatters by the time such long-term investments start yielding a return. The way that GAFAM wield their power today could not be further from the energy sobriety that is called for.

The way that GAFAM wield their power today could not be further from the energy sobriety that is called for.

I spent a large part of my career in the new technology sector. I am familiar with the culture of the Silicon Valley and its giddy faith in the ability of technology to solve all our woes. I can see why it wants to believe that AI may help us deal with climate change.

Unfortunately, the history of new technologies is rather more sobering. The introduction of new, more high-performing technologies has so far failed to help us reduce our carbon footprint. Every time we increase efficiency through innovation, this becomes an opportunity to do more, through a rebound effect, rather than minimise our carbon emissions. And the digital sector’s global emissions keep going up. With AI, this trend has become genuinely alarming.

And the digital sector’s global emissions keep going up. With AI, this trend has become genuinely alarming.

In 2023, 29% of the world’s new wind and solar energy contracts were purchased by GAFAM. Their acquisition of such a high proportion of our carbon-free energy resources creates conflicts of use.

For when such powerful actors choose not to participate in our efforts to achieve energy sobriety, other sectors end up paying the price because they have to make further cuts to their own energy consumption. Do we not agree that generalised AI has no priority over the energy resources available to humankind?

Do we not agree that generalised AI has no priority over the energy resources available to humankind?

We have already broken six (nearly seven) of the nine planetary boundaries within which future generations can continue to thrive on this earth. And we still don’t know how to turn back the clock.

What sort of world do we want to live in? A world where a large chunk of the already stretched capacities of solar panels, wind farms and nuclear plants are the preserve of the digital sector?

A world that fosters soil artificialisation in order to meet the unrelenting demand for new AI data centres? A world in which our digital lives matter more than the natural ecosystems that keep us alive?

A world in which our digital lives matter more than the natural ecosystems that keep us alive?

I strongly believe that we urgently need to come up with a collective plan for frugal innovation. It is no longer possible to think that technology will somehow end up solving the problems that it creates. In fact, I consider that the generalisation of AI is one giant leap in the wrong direction. That this technology should only be developed parsimoniously. That its uses should be clearly defined and in the general interest. And that it should be developed by a non-profit organisation, perhaps under the aegis of an institution like the United Nations.

It is no longer possible to think that technology will somehow end up solving the problems that it creates.

We are not dependent on AI yet. It is not too late to back away. Let us not forget that each and every one of us has the power to inflect the direction that the world takes through the choices that we make as citizens. Let us consider our usage of AI tools like ChatGPT, and their impact on climate change. In view of GAFAM’s soaring carbon emissions, what do we think will happen if our own use of AI helps to further normalise this technology?

Let us not forget that each and every one of us has the power to inflect the direction that the world takes through the choices that we make as citizens.

Personally, I value my digital life much less than our natural ecosystems. As a citizen with a duty to do everything I can to ensure that future generations may inherit a world where life can still thrive, my message is this: Yes, humans are not perfect, but no, we do not need artificial augmentation.