Artificial Intelligence and Energy

Efficient AI rely on models trained on large parameters, that:
– Have high computation/ energy costs and produce CO2
– Are inaccessible except for highly funded funded projects and top tier labs

So:
• Authors should report training time and sensitivity to hyperparameters
• Equitable access to computation resources is required for academic researchers
• Prioritize computationally efficient hardware and algorithms

Full paper: Arxiv

By Vedang R. Vatsa

IT & Management Consultant