During the COVID-19 pandemic, robots helped Carnegie Mellon University students in the Computational Biology Department (CBD) complete lab assignments.

Joshua Kangas, an assistant professor in CBD and co-director of the Master of Science in Automated Science (MSAS) program, said students could log in to the Automated Science Lab's integrated robot, design experimental protocols and use cameras to watch the robots run their experiments.

"We set up the system allowing remote access to the lab, and it was used remotely by students in my courses," Kangas said. "We had students from all over the world using the automation lab."    

This innovative approach to solving lab-access problems was possible because of many prior developments designed to conduct experimentation in a more efficient, cost-effective manner on large, complicated research problems.    

To celebrate this kind of automation and AI-driven science, Carnegie Mellon — long home to leading experts in the fields of AI, automation and the sciences looking to push the boundaries of many disciplines — hosted the third Nobel Turing Challenge Initiative Workshop July 11–12. The workshop was organized by the Nobel Turing Challenge Initiative and sponsored by Carnegie Mellon with additional support from the Systems Biology Institute.

"This workshop holds a special place in our hearts, as the development of AI-driven science has a long and esteemed history here at CMU," said Theresa Mayer, the university's vice president for research. "Work at Carnegie Mellon over many decades helped set the stage for the surge in automated science that we see today."

The workshop brought together an international group of scientists to exchange information and discuss collaborations aimed at further developing highly autonomous AI and robotics systems that can make scientific discoveries, including some that may be worthy of a future Nobel Prize or Turing Award.

The paradigm of automated science, also referred to as autonomous science, self-driving instruments, or self-driving labs, combines laboratory automation (robotic laboratory devices) to execute experiments with AI to interpret the results and decide which experiments to do next. This process is done iteratively, with a goal of either building a predictive model without doing all possible experiments or finding experimental conditions that give a desired result.

Talks at the workshop described six different areas where this approach has been used: microbiology, metabolism modeling, protein engineering, synthetic biology, materials design and organic chemistry. All talks demonstrated the significantly improved efficiency typically achievable using the automated science paradigm.

Robert Murphy, the Ray and Stephanie Lane Professor of Computational Biology Emeritus, led the program committee that organized the workshop. Murphy, who founded the Computational Biology Department, is a pioneer in the field. His recent work focuses on applying and developing active machine learning methods to drive biomedical research campaigns.

"It was a great opportunity for panel sessions and discussions with an eye toward the future," Murphy said. "The two main themes we discussed were: What challenges remain to be addressed with the AI side of driving experimentation? And also: What developments are needed for automated instruments to be better adapted to interface with each other and with AI?"

The Carnegie Mellon Cloud Lab — a shared, remote-controlled, highly automated facility with more than 200 types of scientific equipment — will open later this year. Carnegie Mellon will use the cloud lab platform, developed by the alumni-founded Emerald Cloud Lab (ECL), to accelerate the speed of scientific discovery, spur collaboration in the scientific process, and democratize science by lowering the barriers of cost and availability of scientific equipment for both those inside CMU and the broader scientific community. ECL founders Brian Frezza and D.J. Kleinbaum led tours of the space for workshop participants.

Olexandr Isayev, an associate professor of chemistry at Carnegie Mellon, and his students have been at the forefront of the university's efforts to adopt the Cloud Lab platform, which they started using nearly two years ago. Isayev uses machine learning and AI to develop technologies to rapidly accelerate the pace at which novel molecules, materials or drugs are discovered. He discussed ongoing projects focused on optimizing the properties of polymers for 3D printing in one of the plenary sessions.

"With my students, we develop algorithms to better connect CMU Cloud Lab platforms and experimental instruments. The work creates a feedback loop, where AI executes experiments and teaches itself to make adjustments until desired criteria are met," Isayev said.

While in Pittsburgh, attendees received guided tours of the Cloud Lab and the Automated Science Lab. Conference participants also heard a variety of AI-related scientific talks in the fields of biology, chemistry, materials science and cosmology; participated in a poster session; and had opportunities to network and collaborate on issues related to the field of automated science.

Jose Lugo-Martinez, an assistant professor in the CBD and co-director of MSAS, said that the conversations and presentations were enlightening.

"Automated science is more prevalent than many people realize in many disciplines," Lugo-Martinez said. "It's been interesting to see the progress being made and where there is potential to make enormous progress."

Héctor García Martín, a group lead at the Lawrence Berkeley National Laboratory who works on bioengineering cells to produce biofuels and other renewable products, was also energized by the workshop. As the founder of Berkeley Lab's Quantitative Metabolic Modeling group, he uses synthetic biology, automation and machine learning to make synthetic biology predictable.

"Carnegie Mellon is leading the way in creating the biology of the future," García Martín said.

 View talk recordings on the workshop website.