6 articles
Epoch AI research reveals AI training efficiency improved through compute scaling, reaching 21,400× total gains with estimates of nearly 10× annual improvements driven by scale-dependent algorithmic innovations.
AI training compute efficiency has been improving several times per year, according to Epoch AI analysis. Multiple studies confirm substantial gains - though uncertainty ranges remain wide across different methodologies.
Epoch AI's latest research reveals that AI systems could exhaust the entire supply of public human-generated text by 2028, with over-training issues potentially emerging as early as 2027.
New Epoch AI data shows frontier model performance climbing nearly twice as fast in recent months, though experts caution the acceleration may reflect benchmark optimization rather than across-the-board capability gains.
New Epoch AI data reveals frontier artificial intelligence capabilities started improving almost twice as fast in 2024, driven by advances in reasoning models and reinforcement learning techniques.
Epoch AI's 2025 retrospective reveals exponential acceleration in large language model development, with performance improvements now occurring in months rather than years across capability, efficiency, and research output metrics.
We use cookies to improve your experience on our site and to show you relevant advertising. To find our more, read our privacy policy and cookie policy
Peter Smith
Victoria Bazir
Alex Dudov
Saad Ullah
Eseandre Mordi
Usman Salis
