RESEARCHERS CRACK AI ENERGY CRISIS: NEW SYSTEM USES 100X LESS POWER AND GETS BETTER RESULTS Tufts University researchers built a neuro-symbolic AI that merges neural networks with logical reasoning, achieving a 95% task success rate versus 34% for standard AI — while consuming only 1% of the training energy and 5% of the runtime energy. If scaled, the approach could dramatically shrink the estimated 415 terawatt-hours AI devoured globally in 2024. Keywords: neuro-symbolic AI, AI energy efficiency, AI power consumption, Tufts AI research