From TokenPost · · 50s
Google DeepMind Finds No AI Acceleration in ML R&D
DeepMind's AI research shows no significant speedup in machine learning development despite heavy code automation.
In brief
DeepMind's AI research shows no significant speedup in machine learning development despite heavy code automation. Google DeepMind's internal analysis found no end-to-end acceleration in machine learning R&D despite AI writing 75% of production code. Originally reported by TokenPost.
DeepMind's R&D Assessment
Google DeepMind's internal analysis found no significant end-to-end acceleration in machine learning research and development despite AI writing more than 75% of the lab's production code.
“There is currently no significant end-to-end acceleration of ML R&D, AI-driven or otherwise.”
GCI Index Trend
The GCI v1 index shows frontier Gemini capabilities rose by 0.1 points per month with no significant change after April 1, 2026.
Code Automation Levels
Fully delegated pull requests rose from 7.5% in April 2026 to 11.9% in September 2026, but 88% still needed human steering.
Automation Effectiveness
Automation works best on clearly specified tasks with verifiable outcomes but struggles with open-ended scientific ideas and long-range research.
What to remember
Google DeepMind's internal analysis found no end-to-end acceleration in machine learning R&D despite AI writing 75% of production code.
