Carnegie Mellon University’s AI system, Coscientist, has achieved a groundbreaking milestone in the field of AI-driven scientific research. For the first time, a non-organic intelligent system has successfully automated complex, Nobel-winning chemical reactions, transforming the landscape of lab work. This remarkable achievement was reported by researchers from Carnegie Mellon University in the prestigious journal Nature.
The potential impact of intelligent agent systems for autonomous scientific experimentation is tremendous, according to the Carnegie Mellon research team. They envision the discovery of unforeseen therapies, new materials, and groundbreaking scientific advancements facilitated by the synergetic partnership between humans and machines.
Coscientist, developed by Assistant Professor of Chemistry and Chemical Engineering Gabe Gomes and chemical engineering doctoral students Daniil Boiko and Robert MacKnight, harnesses the power of large language models (LLMs) such as OpenAI’s GPT-4 and Anthropic’s Claude. By receiving simple, plain language prompts, the system can execute the entire experimental process.
Researchers can now ask Coscientist to identify compounds with specific properties. The system scours various sources, including the internet and documentation data, synthesizes the information, and devises a course of experimentation using robotic APIs. The experimental plan is then carried out by automated instruments. Ultimately, researchers working alongside Coscientist can design and conduct experiments with unparalleled speed, accuracy, and efficiency compared to traditional methods.
Coscientist represents a hyper-efficient lab partner, going beyond chemical synthesis tasks to provide genuinely valuable scientific assistance, says National Science Foundation (NSF) Chemistry Division Director David Berkowitz. The system’s ability to seamlessly integrate various components enables scientists to solve optimization problems and complete scientific tasks that rely on multiple hardware modules and extensive data sources.
The use of LLMs in scientific research automation offers numerous benefits. One of the most significant barriers to using automated labs has been the ability to code. However, by enabling scientists to interact with automated platforms using natural language, Coscientist opens doors to a wider range of researchers.
This innovation is transformative, particularly for academic researchers who lack access to advanced scientific research instrumentation commonly found at top-tier universities. The introduction of remote-controlled automated labs, also known as cloud labs or self-driving labs, democratizes science and empowers scientists across various institutions.
To demonstrate Coscientist’s capabilities in an automated robotic lab, the Carnegie Mellon researchers partnered with Ben Kline from Emerald Cloud Lab (ECL), a remotely operated research facility. The successful collaboration showcases not only the value of self-driving experimentation but also a novel means of sharing the research outcomes through cloud lab technology.
In early 2024, Carnegie Mellon, in collaboration with ECL, will open the first university-based cloud lab. The Carnegie Mellon University Cloud Lab will grant researchers and their collaborators access to over 200 pieces of equipment. Gabe Gomes intends to further refine the technologies described in the Nature paper for use in the Carnegie Mellon Cloud Lab as well as other self-driving labs in the future.
Coscientist’s contribution extends beyond laboratory automation. The system ensures every step of the research process is traceable and reproducible, effectively opening the black box of experimentation. By maintaining meticulous documentation, Coscientist enhances the reliability, replicability, and reusability of the generated datasets, benefiting the broader scientific community.
While the potential of AI-enabled science is vast, concerns regarding safety and ethics remain paramount. Gabe Gomes and his team have meticulously addressed these concerns and ensured that the positive impact of these powerful tools outweighs the risks. By incorporating ethical and responsible approaches, scientists can leverage the vast potential of large language models while mitigating possible misuses.
The research conducted by the Carnegie Mellon team has attracted significant attention and recognition. It offers a glimpse into the future of scientific research, where AI-driven systems like Coscientist collaborate seamlessly with human researchers. With the promising developments in automating scientific discovery, the path to groundbreaking discoveries, transformative therapies, and revolutionary materials has been significantly accelerated.
Reference:
Autonomous scientific research capabilities of large language models 20 December 2023, Nature.
DOI: 10.1038/s41586-023-06792-0
This research was supported by Carnegie Mellon University, its Mellon College of Science, College of Engineering, and Departments of Chemistry and Chemical Engineering. Boiko’s graduate studies were supported by the National Science Foundation’s (NSF’s) Center for Chemoenzymatic Synthesis (2221346), and MacKnight’s graduate studies were supported by the NSF’s Center for Computer Assisted Synthesis (2202693).
