Reducing the carbon footprint of technology, the new promise of AI

As data centers, essential for digital life, multiply, the carbon footprint of the technology sector increases, but the artificial intelligence (AI) could reduce it, according to its promoters.

The stakes are high: by 2025 the sector is expected to consume the twenty% of the electricity produced worldwide and is responsible for the 5.5% of total carbon emissions.

In addition, the proliferation of increasingly energy-demanding uses and applications is likely to further accelerate this pace.

“Pandora’s box is open”acknowledged Arun Iyengar, CEO of Untether AI, a company that seeks to manufacture lower-energy semiconductors for AI.

“We can use AI to improve applications and make them compatible with climate requirements, or do nothing and suffer the consequences,” he claimed.

The transformation of the world’s data servers to be AI-ready is underway, a process that a Google executive called “a turning point in computing that happens once in a generation.”

Race for energy efficiency

The development of generative AI tools, such as the GPT-4 chatbot, the basis of ChatGPT’s success, or Google’s Palm2, for Bard, involves two stages, both of which are highly energy-intensive: “training” and execution. .

Researchers at the University of Massachusetts who tested these tools in 2019 found that training a single AI model can account for a similar amount of emissions as five cars over the course of their lifetime.

A more recent study by Google and the University of California Berkeley estimated that GPT-3 training emitted 552 tons of carbon emissions, as much as driving a car for two million kilometers.

OpenAI’s next-generation model, GPT-4, is trained with around 570 times more parameters (or inputs) than its predecessor. The scale of these systems will only grow as AI becomes more powerful and ubiquitous.

At the center of this development are power-hungry graphics processors, or GPUs, made by Nvidia.

Once the training is completed, the use of generative AI tools through the cloud also requires energy through consumption linked to the requests received. And this expense far exceeds that of training.

However, since powerful processors are no longer needed for cloud servers, companies could opt for more environmentally friendly solutions.

Amazon Web Service (AWS), Microsoft and Google, the main cloud players, say they want to reduce their energy consumption.

AWS announced that it aims to be carbon neutral by 2040, while Microsoft aims to be “a company with negative emissions and zero waste” by 2030.

Between 2010 and 2018, data center consumption worldwide increased by only one 6%despite the fact that its use increased a 550%according to the International Energy Agency (IEA).

Global warming

For its promoters, the carbon footprint of AI is not a problem.“When we have really powerful superintelligence, tackling global warming won’t be very difficult”said Sam Altman, founder of OpenAI (ChatGPT).

“This shows how big we must dream. Imagine a system where you can ask ‘tell me how to produce a lot of clean and cheap energy, how to capture carbon efficiently and how to build a factory that can do it on a global scale’”he continued.

The director of Nvidia, Jensen Huang, believes that the massive deployment of AI and the greater speed of computing tools could end up causing a reduction in demand for the cloud and, therefore, in the consumption of the sector.

Thanks to AI, laptops, smartphones and cars could become low-power supercomputers that don’t need to retrieve data from the cloud.

“In the future, you will have a tiny processor in your phone and the 90% of the pixels will be generated, the remaining 10% will be retrieved online, instead of the 100% so you will consume less”Huang told reporters.

But some experts believe the headlong rush to AI is distracting attention from environmental risks.

“Large corporations are currently spending a huge amount to deploy AI. I don’t think they’re worried about the environmental impact yet, but I think that time will come.” Arun Iyengar is excited.