Professor investigates: Can AI and sustainability coexist?
A Professor of Data Science for Sustainability and the Environment at Queen Mary University of London is researching how artificial intelligence can be used to tackle environmental challenges - and asking a crucial question: how can we use AI to promote sustainability while minimizing environmental damage?
It is becoming increasingly clear that the negative impact of AI on the environment is primarily due to its immense energy requirements. Most of this energy consumption occurs in two phases: training the AI models and using them for inference, i.e. the process of generating answers or predictions, e.g. when waiting for an answer to a question.
Training large AI models such as ChatGPT requires enormous computing resources and can run for weeks or months - with correspondingly high energy consumption. Even after training, the demand remains high, as millions of users access the models every day.
This becomes problematic as almost 70 % of global electricity generation still comes from fossil fuels. Experts warn that AI could increase electricity consumption in Europe by up to 50 % over the next ten years - an additional burden on the already challenging transition to clean energy.
Can artificial intelligence help?
Despite its high energy consumption, AI also has enormous potential to drive sustainability forward - especially in the geosciences. AI models are revolutionizing weather forecasting, for example. Programs such as Google's GenCast are already outperforming established models such as the ENS of the European Centre for Medium-Range Weather Forecasts.
While conventional models are slow and computationally intensive, AI delivers more accurate weather forecasts with significantly less energy input. This makes hourly forecasts possible, which could help to detect natural disasters at an early stage and minimize damage.
AI also makes it easier to analyze large amounts of data, such as decades-old satellite images. As a result, deforestation, ocean conditions and the consequences of natural disasters can be monitored more efficiently. AI-supported research also promotes the protection of coral reefs and accelerates the energy transition.
By using AI, researchers gain deeper insights into environmental changes and possible countermeasures. In this way, climate change can be better understood, slowed down or at least made adaptable.
Yes, AI can help
Reducing the environmental impact of AI requires a multi-layered approach. One example of this is the recently modernized physics data center at Queen Mary University of London, which uses waste heat from computer servers to heat buildings on campus. This method shows how data centers - including those used for AI - can be made more sustainable through innovative solutions.
Advances in computer hardware and software are also crucial. New technologies such as quantum transistors could significantly reduce energy losses, while optimized software reduces computing costs. In addition, the water consumption of data centers can be further reduced through improved cooling systems and the use of AI for efficient control.
Ultimately, the sustainability of AI depends heavily on the switch to renewable energies. The expansion of solar, wind and nuclear energy is essential to meet the growing demand for electricity in an environmentally friendly way. Initiatives such as multidisciplinary research centers that promote new ideas for green energy technologies make important contributions in this regard.
Global inequalities in access to AI
Positive sustainability applications of AI have so far mainly benefited the Global North, where infrastructure, funding and expertise are available. This imbalance risks exacerbating existing global inequalities and denying communities in the Global South access to important tools that could contribute to climate adaptation.
Equitable access to AI technologies is therefore crucial. Projects such as improving local weather forecasts in Sierra Leone or training for scientists in Indonesia show how AI can be used to specifically address local environmental problems.
Cooperation between governments, universities and the private sector is needed to drive these developments forward. The focus must be on fair partnerships and knowledge sharing so that AI can deliver sustainable solutions worldwide - even in places where it was previously difficult to access.
Because when it comes to AI and the environment, the challenge is great, but so is the opportunity.
Source: Queen Mary, University of London
