Automatic scientific progress arrives, hand in hand with artificial intelligence

2023-12-22 20:45:03

For the first time, a non-organic intelligent system has designed, planned and executed a chemical experiment.

This has happened thanks to the work of some scientists from Carnegie Mellon University (CMU) in the United States.

“We anticipate that intelligent agent-based systems for autonomous scientific experimentation will bring about important discoveries, including, for example, revolutionary therapies and new materials. Although we cannot predict what those discoveries will be, we hope to see a new way of research derived from the partnership synergy of humans with machines. This is how the creators of the Coscientist system express themselves.

The main members of the team that created Coscientist are Gabe Gomes, Daniil Boiko and Robert MacKnight.

Coscientist uses large language models, such as OpenAI’s ChatGPT and Anthropic’s Claude, to run the full range of the experimental process with a simple plain language instruction.

For example, a scientist can ask Coscientist to find a chemical compound with certain properties. The system searches the Internet and other available sources, extracts data from the documentation it consults, puts the necessary information in order and selects the steps to follow to carry out the experiment, using robotic application programming interfaces. The experimental plan is then sent to the automated instruments, which complete it. A human working with Coscientist can design and execute an experiment much more quickly, accurately, and efficiently than a human alone.

The Cloud Laboratory is a fully automated and remotely controllable Carnegie Mellon University laboratory that allows researchers, whether human or not, to remotely access more than 200 scientific devices with which to perform a wide variety of experiments. It is the ideal scenario for artificial intelligence to go from being an object of study to carrying out scientific research. (Photo: Carnegie Mellon University)

Specifically, the research team demonstrated that Coscientist can plan the chemical synthesis of known compounds; search and browse hardware documentation; use the documentation to execute high-level commands in an automated laboratory (the so-called Cloud Laboratory); control liquid handling instruments; complete scientific tasks that require the use of multiple hardware modules and various data sources; and solve optimization problems by analyzing previously collected data.

Gomes and his colleagues present the technical details of the launch of Coscientist and its pioneering work in the academic journal Nature, under the title “Autonomous scientific research capabilities of large language models.” (Source: NCYT from Amazings)