Cost–benefit of energy-intensive AI training for insight generation

Assess whether the energy and resource costs of training Artificial Intelligence applications are justified by the scientific insights they provide in Artificial Utopia research, and develop metrics to evaluate this trade-off.

Background

The paper emphasizes sustainability concerns related to training AI systems, which can be energy- and resource-intensive. It raises doubts about whether these costs are offset by the insights generated in the context of social simulations for democratic institutions.

This poses a concrete evaluation problem: quantify and assess the trade-offs between computational costs and knowledge gains.

References

Training AI applications is energy and resource-intensive and it is not clear whether the 'costs' are worth the insight \parencite{kaack2022aligning}.

— Artificial Utopia: Simulation and Intelligent Agents for a Democratised Future  (2503.07364 - Oswald, 10 Mar 2025) in Section 4 (Open questions and discussion)

However, the problem of defining 'good work' remains an open challenge.

— Scoring and Gamification to Encourage Sustainable Use of Compute Clusters  (2608.18786 - MacDonald et al., 19 Aug 2026) in Section Open Challenges