Our digital footprint is growing rapidly: the ICT sector already generates more emissions than the aviation industry, and with technologies such as AI models, this trend is expected to continue. That is why, in this edition of “What scientists say about...”, three scientists reflect on sustainable ICT use. “We often overestimate what computers can do, while greatly underestimating what human cognition is capable of.” Jordy Aaldering: “The ICT sector plays an important role in achieving climate targets” “The carbon footprint of the ICT sector is easily overlooked. Its environmental impact is largely invisible -- we don’t see the emissions caused by streaming video, training AI models, or storing data in the cloud. Yet, the ICT sector is already responsible for more greenhouse gas emissions than the aviation industry. If current trends continue, it could account for 14% of global emissions by 2040. Clearly, the ICT sector has a major role in achieving climate goals. One way to reduce this impact is by using the computing resources we already have more efficiently. Recent studies show that in data centres, less than a fifth of available hardware is used, meaning vast amounts of energy are spent powering underutilized machines. To reduce waste, we must write smarter software that better matches workloads to available capacity. Modern computers are composed of many specialized components, designed to excel at specific tasks like 3D rendering or running AI models. While this boosts performance, it also makes it harder for developers to write software that runs efficiently across all hardware combinations. The reality is that software developers today face a dizzying mix of hardware types, as well as unpredictable factors like ambient temperature and other competing processes. Although many developers are eager to write greener code, they often lack the tools and knowledge to do so in practice. That's why we believe the solution lies in automatic and adaptive software systems. These systems adjust how a program runs based on the specific hardware and the real-time conditions. This not only reduces the burden on developers, but also leads to smarter and more flexible software that can optimize itself. By building adaptive energy-aware software, we take a meaningful step toward a more sustainable digital infrastructure that supports our growing reliance on technology.”Jordy Aaldering is a PhD candidate in Software Science. Iris van Rooij: “The ChatGPTs of this world are disastrous for academic education”“Creating artificial general intelligence with human-level cognition is ‘impossible’. Some argue that it is only a matter of time before we have computers that can think like humans. But the principle alone is not enough to make it feasible. Pursuing this goal is a futile endeavour and a waste of raw materials and energy resources.We often overestimate what computers can do, while greatly underestimating what human cognition is capable of. It is important that we help people develop critical AI literacy so that they have the tools to assess how unfeasible the claims of large tech companies are. If a company came along claiming to have a machine that, at the push of a button, would create world peace, you would be suspicious. So why are we so quick to believe the unrealistic promises of large tech companies driven by profit?It is worrying that universities are going along with the hype surrounding the technologies that large tech companies are presenting to us. Every prompt you enter costs a lot of energy, large language models are trained on copyrighted texts, and the underlying models are inaccessible and therefore uncontrollable.But above all, the ChatGPTs of this world are disastrous for academic education. Your brain is shut down. Whereas the university is there to teach you to think independently and critically. Summarising, for example, is an analytical skill that you acquire during an academic education. You view information through your personal cognitive lens, which leads you to different and interesting insights than someone else. That is what science is all about. If you throw everything through ChatGPT, you end up with uniformity.”Iris van Rooij is professor of Computational Cognitive Science. Together with colleagues, Van Rooij published an article in which she further explains why universities should not use AI indiscriminately. Job Doesburg: “Good privacy and security are also a form of sustainability”“Privacy and security are high on the agenda. The measures we take to increase security can be very energy-intensive. By storing less data, processing data more efficiently and deleting data, you reduce the risk of data leaks and work more sustainably.I am a big advocate of deleting data, but once you have deleted data, you can no longer (re)use it. And that data may have been very expensive to collect. That is not sustainable either. With PEP technology (Polymorphic Encryption and Pseudonymisation), developed at Radboud University, we can encrypt and pseudonymise data.This allows us to make maximum use of research data and reuse it, with minimal privacy and security risks. For example, Radboudumc's research into Parkinson's disease involved collecting all kinds of data from patients over many years. With PEP, we have brought all that data together in one large database. Different researchers can request access to only the data they need for various research purposes.Thanks to the security and privacy offered by PEP, researchers can reuse this data, eliminating the need to collect data in all kinds of separate studies. We are now also (re)using this technology at NOLAI (National Education Lab AI) to handle data as sustainably as possible. After all, concerns about privacy and energy consumption in AI are naturally high.Sustainability is about more than reducing your energy consumption. Good privacy and security are also forms of sustainability, because you can reuse data more effectively when privacy and security are guaranteed. And by organising this properly, you can also indirectly improve energy consumption. I see it as a major advantage that we manage our own IT and are not overly dependent on large tech companies. This allows us to maintain control and remain in charge of our data.Job Doesburg is a PhD candidate in Software Science.Sustainability in the MakingRadboud University is working in various ways to make ICT on campus more sustainable. The “Sustainability in the Making” campaign highlights what is already being done and shows what opportunities there are for further sustainability improvements. Read more about the campaignPhoto: Dick van Aalst Contact information Organizational unit Donders Institute for Brain, Cognition and Behaviour, Faculty of Science, Institute for Computing and Information Sciences