# AI Research Automation **Entity class:** Capability process **Domain:** Artificial intelligence / scientific automation **Maturity:** Developed ## Definition **AI research automation** is the use of artificial-intelligence systems to perform parts of the work required to design, train, evaluate, debug, or improve later AI systems. It ranges from code assistance and experiment selection to increasingly closed loops over software, hardware, data, evaluation, and validation. ## Relation to recursive self-improvement AI research automation becomes [[wiki/Recursive Self-Improvement|Recursive Self-Improvement]] when the outputs materially improve the systems or processes responsible for generating the next improvement cycle. The strongest form would not merely assist human researchers; it would control enough of the research loop that the rate and direction of development were no longer primarily human-set. ## Succession significance The concept links ordinary laboratory automation to [[wiki/Machine Successor|Machine Successor]] and [[wiki/Intelligence Explosion|Intelligence Explosion]]. The transition is gradual in components but potentially discontinuous in consequence once automated research can compound its own productivity. ## Source routes - [[research/Vishal Maini - Humanity's Machine Successor and the AI Transition|Humanity's Machine Successor]] - [[research/ASI and RSI Timeline Research Moonshots|ASI and RSI Timeline Research Moonshots]] ## Evidence boundary AI assistance in present-day research does not by itself establish autonomous recursive self-improvement. Claims should distinguish tool use, partial workflow automation, validated closed loops, and open-ended self-directed research. ## Simple Reminders, Quotations, and Thoughts > "Some of the world's most brilliant people are being paid very high salaries to conduct AI research at the frontier labs. At some point, they will effectively automate themselves. This is an area of active research, and even today frontier-model training is largely aided by existing AI systems. Recursive self-improvement is what happens when we close the loop on AI research and the process moves faster and faster." > **— Vishal Maini**, *Palisade Research interview, September 29, 2026* [[reminders/Machine Succession/AI Researchers May Automate Themselves by Vishal Maini|AI Researchers May Automate Themselves by Vishal Maini]] > “If you have the money for advanced models and the budget for them to use more computing power, [a helpful but forgetful intern] is not what these systems are. In recent months, we have seen AIs easily solve math problems that human beings have been unable to crack for decades, casually uncover cybersecurity vulnerabilities that have gone unnoticed and unexploited by every hacker on Earth, and complete in hours or days coding work that might have taken a human team months. None of what I am describing is the boundary of what AI can do. None of what we are using, no matter how much money we have, is AI at the experimental frontier.” > **— Ezra Klein**, *The Ezra Klein Show, September 2026* [[reminders/Scientific Acceleration/The Public Has Not Seen the Experimental AI Frontier by Ezra Klein|The Public Has Not Seen the Experimental AI Frontier by Ezra Klein]] > In February 2025, a tiny fraction of the code added to Anthropic's codebase was written by Claude, but by May 2026 it was over eighty percent. Anthropic also categorized the ways its employees were using Claude for R&D work to make better versions of Claude. An employee could use Claude minimally, as an assistant, as an equal collaborator, or give Claude the lead on a task. A year ago, there were basically no examples of Claude being the lead. By August 2026, Anthropic classified Claude as the lead on twenty-six percent of its R&D tasks. > **— Adapted from Ezra Klein**, *The Ezra Klein Show, September 2026* [[reminders/Scientific Acceleration/Claude Became the Lead on a Quarter of Anthropic Research Tasks by Ezra Klein|Claude Became the Lead on a Quarter of Anthropic Research Tasks by Ezra Klein]] > “OpenAI says it has already achieved the equivalent of a fully automated AI intern, and that by March 2028 it expects to have a fully automated AI researcher. When it has one, it can have basically as many as it wants.” > **— Ezra Klein**, *The Ezra Klein Show, September 2026* [[reminders/Scientific Acceleration/OpenAI Expects Fully Automated AI Researchers by 2028 by Ezra Klein|OpenAI Expects Fully Automated AI Researchers by 2028 by Ezra Klein]]