← Research map
AI for science System technical lens

The AI Scientist: Towards Fully Automated Open-Ended Scientific Discovery

Chris Lu, Cong Lu, Robert Tjarko Lange, Jakob Foerster, Jeff Clune, David Ha

Key signal

A framework for automated scientific discovery in which frontier models generate ideas, write code, run experiments, plot results, write papers, and simulate review.

Open research question

Which stages of the automated research loop remain the dominant bottlenecks to producing reproducible discoveries that survive expert review and independent replication?

Source date
ASI Research note

The AI Scientist is an early full-stack research automation system. Its value is not that it replaces a strong research lab today; it is that it makes the research loop explicit and executable.

Loop

idea = generate_research_idea(literature_context)
experiment = write_experiment_code(idea)
results = run_and_plot(experiment)
paper = draft_paper(idea, results)
score = automated_reviewer(paper)

ASI relevance

ASI depends on accelerating discovery itself. Systems like this make it possible to ask which parts of science are still human-bottlenecked and which can be moved into cheaper, parallel, evaluator-driven loops.