Robot Scientist Adam
In April 2009, researchers at Aberystwyth University and the University of Cambridge published in Science an account of Adam, an automated laboratory system that independently formulated hypotheses about yeast gene function, designed and executed experiments, and interpreted results without human intervention during the cycle.
Background
By the 2000s, biologists studying yeast had identified many genes whose function was still unknown. The organism in question, Saccharomyces cerevisiae, is one of the most studied in all of science, yet gaps remained in the map connecting genes to the enzymes they encode. Filling those gaps by conventional means was slow. A researcher would form a hypothesis, design an experiment, wait for results, revise, and repeat. Each cycle took time and human attention.
Automation had already changed parts of laboratory work. Robotic equipment could run assays and handle samples faster than any person. But the machines of that era did what they were told. They did not decide what to test next, and they could not look at a result and work out what it meant. The thinking still happened in a person’s head. That boundary between automated execution and genuine reasoning had not been crossed.
Computational approaches to biology were also advancing. Inductive logic programming, a method for learning rules expressed in formal logic from examples and background knowledge, had been used to draw conclusions from biological data. Active learning, where a system chooses its own next query to get the most useful information as quickly as possible, was an established idea in machine learning. Nobody had yet built something that put those ideas together with a physical laboratory and let the whole thing run.
What happened
Ross King at Aberystwyth University, working with colleagues including Kenneth Whelan, Felix Jones, Julian Wales, Andrew Clare, Mark Untill, Stephen H. Muggleton and Douglas B. Kell, built a system they called Adam. The paper describing it appeared in Science on 3 April 2009 (Vol. 324, Issue 5923, pp. 85–89).
Adam was connected to the physical equipment of a real laboratory. It used inductive logic programming to form hypotheses about which genes in Saccharomyces cerevisiae encoded particular orphan enzymes, enzymes whose genetic origins were not yet known. It then used active learning to decide which experiments would be most informative, instructed the robotic hardware to carry them out, collected the results, and interpreted them, all without a human making decisions at any point in that cycle. When one round finished, Adam used what it had learned to plan the next round. The team verified Adam’s conclusions independently and confirmed that several of its discoveries about gene-to-enzyme relationships were correct.
What made this different from earlier automated science was the closure of the loop. Previous systems might automate one part of the process: the bench work, or the data analysis, or the hypothesis generation. Adam did all of them in sequence, with each stage feeding the next. The computer scientists and biologists at Aberystwyth and Cambridge had not built a faster pipette. They had built something that could, within a defined domain, decide what it wanted to know and then go and find out.
Why it mattered
Adam was among the first documented systems to close the full scientific discovery loop autonomously (hypothesis generation, experimental design, physical execution, and result interpretation) without human involvement at each step. This demonstrated that machine learning and laboratory robotics could be integrated to conduct genuine scientific inquiry, not merely assist human researchers. The work established a proof of concept for autonomous science that influenced subsequent research into AI-driven drug discovery and biological experimentation.
People
Ross King, Kenneth E Whelan, Felix Jones, Julian Wales, Andrew Clare, Mark Until, Stephen H Muggleton, Douglas B Kell
Organisations
Aberystwyth University, University of Cambridge
Sources
- The Automation of Science.Science (AAAS).Primary source
- Robot Scientist Project, Aberystwyth University Computer Science.Aberystwyth University.Official
- Robot scientist makes gene function discoveries.University of Cambridge Research News.Secondary
Cite this page
AI Achievements. (2009). Robot Scientist Adam. Retrieved 2026-08-22, from https://achievements.ai/milestone/the-scientist-robot-adam
@misc{achievements_the_scientist_robot_adam,
title = {Robot Scientist Adam},
author = {{AI Achievements}},
year = {2009},
url = {https://achievements.ai/milestone/the-scientist-robot-adam}
}