Suffolk Reporter

Curious minds gathered at Napper Tandy’s in Bay Shore, New York, on November 19, 2024, for the latest installment of PubSci, a science café presented by the U.S. Department of Energy’s Brookhaven National Laboratory. The event aims to connect scientists and the community for informal discussions about the lab's research. This time, the focus was on artificial intelligence (AI) with the topic "Reimagining Scientific Discovery with AI."

At Brookhaven Lab, researchers are exploring how AI can be used as a tool in scientific research. Applications include materials discovery, autonomous experimentation, data processing, brainstorming bots, and virtual research assistants.

The evening began with a toast to those gathered to discuss this frontier in science. Audience members engaged three Brookhaven Lab experts with questions: Carlos Soto from the AI Theory and Security Group; Kevin Yager from the Electronic Nanomaterials group at the Center for Functional Nanomaterials (CFN); and Esther Tsai, a staff scientist at CFN.

Among many questions raised were concerns about the reliability of AI-collected information, development of AI models for science, time savings provided by AI technology to researchers, and whether bots can self-teach. The panel drew comparisons between scientific AI tools and everyday tools like Google Maps and virtual assistants but highlighted significant differences in their application.

“We are all in the space of trying to adapt AI methods to science,” Yager said. He explained that one goal is adapting AI methods to leverage trusted scientific artifacts such as vetted publications or software.

Yager shared a vision for integrating different AI models into a single ecosystem of agents that will accelerate scientific workflows.

Tsai is working on generative AI to create an assistant capable of operating complex instruments on behalf of researchers but stressed human involvement remains crucial. “We should still try to learn things and be curious,” Tsai said.

Brookhaven scientists have also applied AI methods for autonomous experiments where AI directs advanced instruments and adapts dynamically. This technology recently led to discovering new nanostructures through rigorous success determination methods.

Machine learning helps handle vast amounts of data under the broader umbrella of AI. It aids scientists in exploring patterns within data that might be challenging for humans alone.

“Most of AI in science doesn’t look anything like ChatGPT,” Soto noted. He added that data compression and filtering with AI is highly effective.

This capability is particularly relevant for facilities like Brookhaven's Scientific Data and Computing Center (SDCC), which recently stored over 300 petabytes of critical experimental data from major physics research facilities.

“We see AI as another tool and another resource for getting more value out of that data we already have,” Soto stated.

As discussions concluded, attendees heard directly from scientists about working in this rapidly evolving field. Tsai emphasized human responsibility in using AI technology wisely: “I’m very excited about having this great tool that can really accelerate science discovery.”

Yager described his experience as exhilarating: “Sometimes we build solutions that use a model; then a new model comes out better... everything just gets better.”

Soto agreed: “The rate of progress...is incredibly exciting.”

Since its inception in 2014, PubSci has offered Long Island residents insights into cutting-edge research at Brookhaven Lab while allowing casual interactions with scientists over drinks across various topics ranging from cosmology to future technologies.