ORNL's Dr. Rob Moore, an expert in the development of autonomous science and self-driving laboratories, has been discussing how artificial intelligence and automation have led to the cutting edge of autonomous scientific research. In Part 1, Moore explored the origins of AI in science and the challenges of the lab's INTERSECT initiative, along with what's needed for autonomous science to reach its full potential. In Part 2, he examines the changes required as science integrates AI and looks ahead to the future of autonomous science.

Q. You mentioned the hard questions we are trying to answer. Is AI the key to those answers?

A. There are a lot of emerging societal challenges that scientists need to address, and AI can help with these. A lot of these have been there for decades - some of them longer than my career. They're just so hard that we either can't solve them or we're just not doing it fast enough. I think we can solve them, but it's just a matter of when. Some of these have a lot of societal impact. We can't just sit back and wait for solutions to emerge. This is where we can use AI tools to discover faster.

There are very steep challenges and hurdles that we have to overcome to make it happen, however. We're trading one set of challenges for another set of challenges, but the second set of (AI-related) challenges are much more tractable. If we can overcome those, it will be an accelerated pathway to solve the first set (of challenges).

Q. Now you're moving INTERSECT toward the Labs of the Future initiative. How will that be different?

A. The Labs of the Future initiative builds upon INTERSECT, but they are different. INTERSECT focused on developing an ecosystem where researchers could easily connect different resources across the lab. There were some autonomous workflows developed during the INTERSECT initiative, but they focused on an isolated set of instruments and computer resources.

For Labs of the Future, we will utilize this ecosystem, but the focus is to develop larger-scale, closed-loop, autonomous workflows that will connect multiple instruments, labs, and facilities across multiple directorates at ORNL. While INTERSECT focused on developing capabilities, LOTF will focus on developing autonomous workflows and delivering science. This effort will help ensure we understand how to prepare the laboratory for any future Genesis Mission efforts.

Q. How do you think this push will impact scientists in the future? Will AI/automation push humans out of the loop?

A. No. I do not believe that AI will push scientists out of the loop, but it will change how we do science. Science is about research, and if something is real, it is reproducible. This means that as scientists, we often do repetitive tasks many times to ensure we understand what is real and what is an artifact. These tasks may have subtle changes from run to run, but science is a field with lots of repetition.

With AI, we will be able to offload many of the repetitive tasks to agents. The agents can perform repetitive tasks very precisely. They don't get tired and don't need sleep, and they can prepare the resulting outcomes of the tasks in a way that is easy to ingest by the human researchers. This frees up significant researcher time, and they can spend more time thinking about all the information before them and how it fits together to create a bigger picture.

Instead of spending significant time and effort "turning the crank" on a particular measurement or calculation, researchers can take a step back and think more holistically and strategically, things that humans are really good at. There will still be tasks that humans need to perform, but they will be different.

Q. Will AI and automation impact every field of science or are there some areas more prone to it than others?

A. I think that AI will ultimately affect all fields of science, but some will be affected sooner than others. This isn't because AI is better at some scientific domains than others, but because some fields are AI ready and some fields are still working to become AI ready. There are still many fields where critical data is stored in formats that are not easily accessible to AI. It will take time to make these files ready for AI driven science.

Q. Give us a quick sketch of what you think science will look like in 10 or 20 years? How integrated will AI be? What role will humans play?

A. In 10 or 20 years, I envision a lab where researchers arrive to work only to see that the work never stopped. AI agents and robotic tools assist researchers in developing new ideas, designing and executing new experiments, compiling and analyzing new information, and generating results that can be used by the larger community. Humans will spend more of their time collaborating with colleagues and generating new ideas versus performing mundane and repetitive tasks in the lab. Agents will continuously monitor the systems in the lab and alert humans if something is going wrong. The human can then work with the agents to assess the problem, connect with the right expertise to solve the problem and implement solutions through instant knowledge transfer. Humans will spend less time managing the operations of the scientific enterprise and more time thinking holistically about the information being generated and use their intuition, collaborating with other humans as well as AI agents, to think more creatively about scientific problems, as well as possible ways to solve them.

Read Part 1 and Part 2 to explore Moore's discussion of AI's origins in science, the challenges of the INTERSECT initiative, the changes needed as AI integrates into scientific practice and his vision for the future of autonomous science.

UT-Battelle manages ORNL for DOE's Office of Science, the single largest supporter of basic research in the physical sciences in the United States. The Office of Science is working to address some of the most pressing challenges of our time. For more information, please visit energy.gov/science. - Greg Cunningham