Artificial intelligence is going to change our daily lives and the entire world: from business management to medicine, our way of working, learning, and even entertainment. It promises to transform our lives and industries on an unprecedented scale.
But as this technology develops at lightning speed, one question is coming up more and more frequently: does AI pollute? Is artificial intelligence polluting, and what exactly is its carbon footprint?
While the answer seems obvious, it actually holds many surprises.
In this article, we will explore the environmental impacts of AI, while discovering how this technology could very well play a key role in reducing pollution.
AI, an energy-intensive industry that pollutes
Before answering the question, it’s important to understand that behind every artificial intelligence program lies a massive IT infrastructure.
But how does AI pollute?
These systems run thanks to data centers that house thousands, or even millions, of servers. Tech giants like Google, Amazon, and Microsoft are known for their immense infrastructures that require a staggering amount of energy to operate 24/7.
Every time we ask an AI to solve a complex problem or process large amounts of data, it calls upon powerful algorithms that require a lot of computing power.
Model training: an unsuspected pollution
The process of training AI models is particularly energy-hungry. For example, training a natural language processing model (like ChatGPT) can consume as much energy as a car over its entire lifetime, fuel included.
A 2019 study by the University of Massachusetts Amherst revealed that training a single natural language processing model can emit more than 284 tons of CO2, the equivalent of five American cars over their entire lifetime.
This aspect of AI, little known to the general public, clearly illustrates the environmental cost hidden behind this technology.
Why is AI energy-intensive?
One of the main reasons for AI’s high energy consumption lies in the way it learns. Artificial intelligence systems, and particularly those based on deep learning, must process huge volumes of data to improve their accuracy.
This process involves continuous iterations, where the system adjusts its predictions based on previous results. This requires an enormous amount of calculations which, in turn, consume a lot of energy.
Data centers, a key factor
Data centers, which house the servers needed for these calculations, require sophisticated cooling systems to prevent overheating. These systems, in turn, consume a significant amount of energy.
According to a report by the International Energy Agency, data centers account for about 1% of global electricity demand, a proportion that could climb as AI continues to develop.
See also: Here are the legal professions that can be replaced by AI
AI and digital pollution: a complex question
AI pollutes, certainly, but it’s essential not to oversimplify the debate.
If we look at all digital technologies, of which AI is a part, they are responsible for about 4% of global greenhouse gas emissions, a figure that could double by 2025.
This includes the production of IT equipment, the electricity consumption of data centers, and the management of electronic waste.
However, these figures must be put into perspective, as the environmental cost of AI systems is only a small part of the pollution generated by the digital economy as a whole.
The impact of AI on the environment
Although artificial intelligence may seem like a significant source of pollution, the numbers tell a different story.
In reality, AI is far from being one of the largest contributors to pollution compared to other sectors.
For example, the energy industry, which uses fossil fuels to produce electricity and heat buildings, is responsible for about 25% of global greenhouse gas emissions.
Agriculture and deforestation contribute about 24%, notably due to intensive livestock farming and unsustainable agricultural practices.
Furthermore, manufacturing and construction industries, such as the production of steel, cement, or chemicals, generate about 21% of global emissions.
AI: A minimal impact on the environment?
Here is the list of the main sources of pollution on Earth, ranked in descending order of their approximate contribution:
-
- Energy industry (electricity production and heating): ~25%
- Agriculture, forestry, and other land uses: ~24%
- Manufacturing and construction: ~21%
- Transport (road, air, sea): ~14%
- Residential and commercial buildings: ~6%
- Food production (animal feed, fertilizers, transport): ~5%
- Digital technologies (including AI, data centers, internet, electronic devices): ~4%
- Waste management (landfills, incineration): ~3%
- Mining and natural resource exploitation: ~2%
- Plastic waste and ocean pollution: ~1%
Artificial intelligence contributes to pollution, certainly, but much less than one might think. Isolating the impact of AI compared to digital technologies as a whole (which represents about 4% of global greenhouse gas emissions) is difficult, as AI uses all digital components.
However, estimates suggest that AI could represent about 1 to 2% of this total.
This means that AI, while it has an impact, remains a relatively small fraction compared to all digital technologies.
Read also: Is AI-generated text plagiarism? Here is the answer
Solutions to reduce AI’s ecological footprint
Faced with these findings, players in the tech sector are beginning to realize their environmental responsibility. Here are some ways to reduce the impact of AI on the environment.
Algorithm optimization
One of the most promising solutions for reducing AI’s energy consumption lies in algorithm optimization. This involves improving machine learning techniques so they require fewer calculations, and therefore less energy.
Researchers are already working on so-called “lighter” models that require fewer resources while remaining just as powerful. This is a key area of research that could limit the ecological impact of AI systems.
The use of renewable energy
Major cloud computing players, like Google and Microsoft, have committed to using more and more renewable energy to power their data centers.
Google, for example, claims to have been running entirely on green energy since 2017. These companies are also investing in technologies that make their data centers more energy-efficient.
Reusing pre-trained models
Another approach is to reuse already trained AI models. Instead of training a model from scratch, it’s possible to use an existing model that has already learned on a similar dataset and adapt it to a specific task. This saves a significant amount of resources while still achieving high-quality results.
See also: Discover the best AI for generating images
AI, an ally in the fight against pollution?
Paradoxically, AI could also play a major role in reducing pollution. Many AI-based technologies and systems are being developed to help fight climate change and reduce carbon footprints. Let’s see how AI could become an ecological asset.
Energy optimization in buildings
AI can be used to reduce energy consumption in buildings, which account for about 40% of global energy consumption. Thanks to smart sensors and automated management systems, buildings can adjust their consumption based on actual needs.
For example, heating, ventilation, and air conditioning systems can be controlled by AI algorithms that adjust settings based on room occupancy, outdoor temperature, and other factors.
Power grid management
AI is also used to optimize power grid management. Smart grids can use AI systems to adjust energy production and consumption in real time. This not only reduces energy losses but also integrates renewable energy sources, like wind and solar, which are intermittent by nature, more effectively.
Sustainable agriculture
In the field of agriculture, AI could help reduce the carbon footprint of this sector by optimizing the use of natural resources. For example, drones and sensors can collect data on fields, while AI algorithms analyze this information to adjust the amount of water, fertilizer, or pesticides needed.
This approach, known as precision agriculture, reduces agricultural inputs and, consequently, the pollution of soil and groundwater.
Waste reduction and logistics
AI can also play a role in waste reduction by optimizing supply chains. Smart systems can predict consumer demand and adjust production accordingly, thereby reducing waste.
Additionally, AI-based technologies are already being used to improve transport and logistics efficiency, helping to reduce CO2 emissions in the freight transport sector.
See also: Artificial Intelligence: Here is a very simple definition to understand easily
AI can clean up: Improbable but fascinating scenarios
While AI can reduce pollution in concrete ways, some more futuristic and improbable theories are worth exploring.
Autonomous AIs in space
Imagine a future where AI is used to exploit space resources. Asteroids are full of rare metals that are essential for manufacturing clean technologies, like solar panels or batteries.
By sending autonomous robots equipped with artificial intelligence into space, it would be possible to mine these resources without directly impacting the planet.
AI to restore ecosystems
Another scenario envisions using AI to restore damaged ecosystems. Autonomous drones, equipped with AI algorithms, could be deployed to plant trees in areas affected by deforestation.
These drones would be capable of detecting the most suitable areas for planting and acting autonomously to restore entire forests.
Conclusion: a complex balance
So, does AI pollute? The answer is yes, but not in as simple a way as one might think. Pollution related to AI is very real, due to the high energy consumption of data centers and algorithm learning processes. However, AI also presents enormous potential for reducing pollution in other sectors, notably energy, agriculture, and logistics.
The future of AI will likely be marked by a tension between these two aspects. On one hand, it will be crucial to develop more energy-efficient AI technologies. On the other, AI could become a key player in the ecological transition, by optimizing the use of natural resources and contributing to the reduction of greenhouse gas emissions.
Could artificial intelligence one day cancel out its own environmental impact? That remains to be seen, but one thing is certain: AI will continue to surprise us, in its challenges as well as its solutions.
