Rethinking Computers for the Era of Artificial Intelligence

The insatiable energy demands of artificial intelligence (AI) are posing significant challenges to energy systems globally. As companies vie for market dominance in the AI sector, there arises a tension in which increasingly larger AI models necessitate faster data centers that require more resources and new power generation facilities. An alternative approach is to adopt a long-term perspective and develop new architectural designs for computing systems.

Current computing systems are largely ill-suited for the calculations that AI demands. Presently, data centers consume vast amounts of electricity, predominantly sourced from non-renewable energy. Cooling these systems often necessitates substantial water use, which can be particularly problematic in arid regions. Research is underway at the Institute of Computer Engineering at the Technical University of Vienna to conceptualize future data centers that can operate without water for cooling and significantly reduce their energy consumption.

Europe is poised to take a distinct path in the AI landscape. Unlike the United States, where a handful of companies dominate the market, Europe is fostering a more diverse ecosystem. Initiatives like "AI Factories" are being encouraged by the EU, enabling large data centers to be utilized by smaller companies for their AI projects. This initiative aims to create an ecosystem that is accessible to businesses, researchers, NGOs, and more, seeking to reduce reliance on major U.S. corporations.

The EU's energy efficiency directive mandates that data centers adhere to specific efficiency standards and reporting requirements. For instance, a growing percentage of the electricity consumed must come from renewable sources, and waste heat must be harnessed effectively.

Austria is recognized for its strong capabilities in quantum computing and computer science, which presents considerable potential for advancement. However, investment in research is crucial, and strategic planning is needed to build upon these strengths. Currently, there exists an opportunity to attract top researchers from the U.S. to Europe, provided that career paths in research within Austria are made more appealing. There are currently insufficient positions available for leading researchers in the country.

Regarding the EU's climate objectives, there is a dual challenge. On one hand, the high resource consumption associated with AI contributes to greenhouse gas emissions; on the other hand, if utilized effectively, AI could also assist in resource conservation. The key question is whether the advantages of employing AI will eventually surpass its drawbacks. Manufacturers are currently under significant pressure to develop more efficient hardware for AI applications. This situation raises hopes for disruptive innovations that could transform the manner in which software operates.

In terms of public engagement with AI, the notion of individuals forgoing its use for sustainability reasons is dismissed. The broader economic implications must be considered, as individual actions alone would not lead to substantial change. However, enhancing education about artificial intelligence is deemed essential. The current absence of foundational computer science education and logical reasoning in Austrian schools is concerning. It is vital to teach students about the workings of AI and the underlying models. Failing to do so means remaining perpetually as mere users of technology.