On October 1, at a summit on artificial intelligence in Washington DC, Eric Schmidt, former CEO of Google, stated that the energy demand for generative artificial intelligence (AI) will be so enormous that we will never achieve our climate goals. so what we need to do is let AI itself solve the problem of climate change. It is a typical case of so-called techno-solutionism: let AI solve the problems that it itself generates.
Literally, Schmidt said: “My personal opinion is that we will not meet the climate goals because we are not organized to do so, and yes, the needs of AI in this area will be a problem, but I would rather bet on AI solving the problem rather than to restrict it.” Obviously, Schmidt has a significant stake in companies working on the development of generative AI, and therefore, there is clearly a conflict of interest.
The escalation of AI has a direct impact on energy consumption
Schmidt assumes that the future benefits of AI will offset the environmental costs it generates. However, there is no solid argument to support this assumption.
Almost all promises about the great benefits of generative AI are just speculation. We have been told that it will help make great scientific discoveries in fields such as physics, medicine or materials technology, but so far the results obtained thanks to generative AI have turned out to be irreproducible and exaggerated, and in some cases, they have been refuted by subsequent scientific studies.
There is no strong argument that the future benefits of AI will outweigh its environmental costs.
On the other hand, disadvantages, such as misinformation, non-consensual pornography or scams with deepfakesUnfortunately, they are very real. Trusting that generative AI, which has serious problems reasoning in complex or rare situations, can solve climate change is a utopia. It is too risky to bet on this technology at this time.
What is clear is that AI scaling—the strategy of creating larger and larger AI models—has a direct impact on energy consumption. The larger these models are, the more power they require. For example, Microsoft wants to reactivate the Three Mile Island nuclear plant, which in 1979 suffered the worst nuclear accident in US history, to meet the energy needs of its AI.
Generative AI has serious problems reasoning in complex or rare situations
This type of scenario will further aggravate the damage to the environment. It is the typical pseudo-logical mentality of Silicon Valley: pretending that technology is magical and, therefore, its costs do not matter. It is the mentality that allows big technology companies to continue attracting multi-million dollar investments and enrich themselves at the expense of the environment and collective well-being.
There are other more specific and specialized types of AI that could be useful in combating climate change, but these are not the ones that are receiving large investments. Betting on generative AI as a solution to climate change is, as AI and sustainability expert Thomas Dietterich says, absurd. Diversifying efforts in the fight against climate change by directing funding towards much more environmentally friendly specialized AI would make much more sense.
Big technology companies enrich themselves at the expense of the environment and collective well-being
We should not make technological bets without a transparent and responsible process. We need independent studies that evaluate which technology is best, what the chances of success are and what the associated costs are. These decisions should not depend exclusively on a group of billionaires and for-profit companies. Society as a whole must be able to decide, because we are risking the future of our planet.
Ramon López de Mántaras is a professor at the Artificial Intelligence Research Institute of the CSIC
#artificial #intelligence #provide #solutions #climate #change #Ramón #López #Mántaras
