Khaberni - Tech expert Warda Baylish Barshad confirmed that generative artificial intelligence is no longer just a digital concept as it is marketed, but has become a massive physical infrastructure that rapidly consumes energy, water, and natural resources.
Barshad warned that employing artificial intelligence without strategic awareness could turn it into a factor that exacerbates the climate crisis instead of being part of the solution.
In an exclusive interview, Barshad, the head of the international consulting firm BAULD specializing in data and artificial intelligence and author of several books, among them "The Human in the Age of Artificial Intelligence," and an international lecturer on data and ethical artificial intelligence strategies, and an expert in digital transformation, stressed the need to dismantle the "myth of digital immateriality". She considers that what is called "the cloud" actually relies on gigantic data centers, submarine cables, and servers that consume tremendous resources.
How do you see the real environmental impact of artificial intelligence?
To understand the real impact, we first need to break away from the myth of "immateriality," because artificial intelligence is not a virtual entity, but a huge physical infrastructure. Every line of code and every generated image has a real body, which is servers, data centers, and cables that consume energy and exhaust resources. From my position as a strategist, I see it as a complete industrial system, not just an abstract idea.
Where is the greatest energy consumption in the life cycle of artificial intelligence?
There are clearly two stages. The first stage is "waking" or training, which is the most intense in terms of energy consumption, requiring thousands of computational operations for months. For instance, training a model the size of GPT-3 produced carbon emissions equivalent to hundreds of flights between Paris and New York.
The second stage is the service or inference stage, where the challenge becomes systematic, as every daily use accumulates ongoing energy consumption. Each request directed to generative artificial intelligence consumes energy surpassing a traditional internet search by about tenfold, which I call "cognitive energy inflation."
Can the environmental footprint of artificial intelligence be measured accurately?
Yes, it can be measured, but a more important question is: what do we choose to measure? Currently, data centers consume between 1 and 2% of global electricity. However, this figure is insufficient. I work according to what I call the "trinity of impact": the carbon footprint, which is visible and can be measured accurately, and the water footprint, which is a blind spot, as data centers require huge amounts of water for cooling. A single conversation of 20 to 50 questions with artificial intelligence could equal the consumption of half a liter of water. And the physical footprint, which includes rare metals and the short lifespan of equipment due to the rapid obsolescence of AI chips.
What sectors are most affected by these transformations?
There are three forefront sectors. Firstly, the energy sector, which faces unprecedented demand for electricity, and secondly, the technology sector, which is forced to shift from a race for power to a race for efficiency. Thirdly, the service sector, especially finance and media, where artificial intelligence has become the largest element in the digital carbon footprint. But the risk is not only environmental, it also extends to the cultural and social dimensions, with the proliferation of artificial intelligence models lacking local cultural roots.
How can a balance between innovation and sustainability be achieved?
The balance starts with a clear strategy and a triple audit covering carbon, water, and rare metals, prioritizing small, specialized, locally hosted language models, and incorporating "green" clauses in cloud computing contracts to set an energy consumption cap. Implementing circular solutions, such as reusing the heat from data centers to heat cities, and enhancing transparency and accountability through "technology for good" charters.
How do you see the future of artificial intelligence in the context of climate change?
Artificial intelligence is a multiplier. If managed solely based on performance logic, it will exacerbate the climate crisis. However, if directed with ethical intelligence and strategic sense, it can become a lever for the solution, even a "immune system" for the planet, through computing infrastructures aware of carbon and synchronized with renewable energy availability.
What is the most important advice for decision-makers?
I recommend moving from a logic of speed to a logic of precision. Blind innovation is a heavy debt on future generations. I also urge governments to adopt audits for cultural sovereignty, to protect communities from identity-less AI models that empty local knowledge of its meaning.
