AI scientist
In plain wordsAn AI system meant to do research largely on its own: reading the literature, coming up with ideas, running experiments and writing up the results.
An AI system designed to carry out the research cycle with little human help: surveying literature, generating ideas, designing and running experiments (in code or in automated laboratories), analyzing results and writing papers. Sakana AI's 'The AI Scientist' (2024) produced complete machine-learning papers for under US$15 each and scored them with an automated reviewer, and in 2025 a paper written entirely by its successor passed peer review at an ICLR workshop before being withdrawn as planned. Evaluations find such systems prone to fabricated or overstated results, weak checks on novelty and subtle experimental errors, and journals and conferences have debated whether they will accelerate science or flood it. AI scientists working on AI itself are central to scenarios of automated AI researchautomated AI research, n.: Using AI systems to do the work of AI researchers, such as running experiments and writing code for new models, so that AI helps build its own successors. and recursive self-improvementrecursive self-improvement, n.: An AI improving its own design or training, so that each improved version is better at making the next improvement. It is the engine usually imagined behind an intelligence explosion..
The AI scientist ran 400 ablations overnight, but a human still had to spot the data leak.
Origin
The ambition dates back to DENDRALDENDRAL, prop. n.: A program begun at Stanford in 1965 that worked out the structure of chemical molecules from laboratory measurements using rules taken from expert chemists. It is usually called the first expert system. (1965) and the robot scientist 'Adam', which Ross King's team reported in 2009 had autonomously generated and tested hypotheses in yeast genetics. The phrase took its current sense with 'The AI Scientist' by Chris Lu, Cong Lu, Robert Tjarko Lange, Jakob Foerster, Jeff Clune and David Ha, posted to arXiv on 12 August 2024.
Related
- AI co-scientistAI co-scientist, n.: An AI system built to work alongside human scientists, proposing and debating new research ideas and hypotheses that people then test in the lab.
- automated AI researchautomated AI research, n.: Using AI systems to do the work of AI researchers, such as running experiments and writing code for new models, so that AI helps build its own successors.
- AlphaEvolveAlphaEvolve, prop. n.: A Google DeepMind system that combines AI language models with an evolution-like search to invent and improve computer programs, finding new results in mathematics and computing.
- Darwin Gödel MachineDarwin Gödel Machine, prop. n.: An experimental AI coding assistant that improves itself by rewriting its own code, testing each new version and keeping a family tree of variants to build on.
- deep researchdeep research, n.: An AI feature that spends minutes rather than seconds on a question — planning, running dozens of web searches, reading many sources — and then writes a long report with citations.
- Genesis MissionGenesis Mission, prop. n.: A US government program launched in 2025 to speed up science with AI, using the national laboratories' supercomputers and data to build AI tools for research.
- AI co-scientistAI co-scientist, n.: An AI system built to work alongside human scientists, proposing and debating new research ideas and hypotheses that people then test in the lab.
Sources
- Lu, C., Lu, C., Lange, R. T., Foerster, J., Clune, J. & Ha, D. (2024). The AI Scientist: Towards Fully Automated Open-Ended Scientific Discovery. arXiv:2408.06292. arxiv.org/abs/2408.06292
- King, R. D. et al. (2009). The Automation of Science. Science 324(5923).
Mentioned in
automated AI researchautomated AI research, n.: Using AI systems to do the work of AI researchers, such as running experiments and writing code for new models, so that AI helps build its own successors., AI co-scientistAI co-scientist, n.: An AI system built to work alongside human scientists, proposing and debating new research ideas and hypotheses that people then test in the lab., AlphaEvolveAlphaEvolve, prop. n.: A Google DeepMind system that combines AI language models with an evolution-like search to invent and improve computer programs, finding new results in mathematics and computing., Genesis MissionGenesis Mission, prop. n.: A US government program launched in 2025 to speed up science with AI, using the national laboratories' supercomputers and data to build AI tools for research., research tasteresearch taste, n.: The knack good researchers have for choosing worthwhile problems, designing the right experiments and drawing the right conclusions — a skill seen as key to whether AI can do research on its own., symbolic regressionsymbolic regression, n.: Getting a computer to discover a readable mathematical formula, such as a law of physics, that fits a set of measurements, instead of only fitting numbers into a formula chosen in advance.