A Rutgers physicist working with Claude on a research project says universities must prepare students for a new model of scientific research
For years, Rutgers physicist David Shih solved Rubik’s Cubes with his children, twisting the colorful squares until the puzzle was back in order.
He didn’t expect the toy to connect with his research, but he recently realized that the logic behind the puzzle was exactly what he needed to solve a particle physics problem.
That idea led to a new artificial intelligence (AI) method that can simplify some of the extremely complex equations used in particle physics. Shih described the method in a study posted on arXiv, a widely used site where scientists share new research.
“In arriving at our solutions, we found that an analogy between mathematical simplification and solving Rubik’s Cubes was key,” says Shih, a professor at the University of California. Department of Physics and Astronomy at Rutgers School of Arts and Sciences. “Both can be seen as problems with encryption and decryption.”
The project also led to something else: Shih conducted the research in full collaboration with an AI system that helped write code, run experiments and produce the paper. This offered a glimpse of a new model of scientific research in which scientists collaborate with AI systems that help design programs, analyze data, and test ideas. It also raises new questions about how to train graduate and undergraduate students for this type of AI-enabled research.
“This research is also notable for the way it was conducted in full collaboration with Claude Code, an agentic AI system that did all the practical work under my supervision,” said Shih, also a professor in the New High Energy Theory Center.
Jack Hughesa distinguished professor and chair of the Department of Physics and Astronomy, said the work highlights how quickly research is changing.
“This new style of research, done in collaboration with AI agents, has the potential to greatly accelerate our research,” said Hughes, an astrophysicist. “There is an urgent need to train our students and postdocs in this new style of research.”
Shih said working with AI changed what he could try as a researcher.
“If we learn how to use these tools properly, we will be able to tackle more ambitious problems,” he said. “It changes the scale of what one person can do.”
In many areas of science and engineering, equations can become extremely long and complicated. In particle physics, equations describing subatomic particle collisions can contain hundreds of terms, but physicists insist that beneath the complexity lies something simple and elegant.
“Particle physics in particular is a field where complex calculations are often expected to lead to simple answers because of the symmetries and structures underlying the fundamental theory,” says Shih.
By simplifying these equations, scientists can see patterns more clearly, make more accurate predictions, and reduce the processing power required for calculations. Simplifying an equation does not change its meaning, but it can make calculations more accurate by avoiding the subtraction or combination of very large numbers that can cause small rounding errors in computers.
Shih wanted to see if AI could help find new ways to simplify these equations using what’s known as machine learning, a type of artificial intelligence in which computers learn by studying examples and recognizing patterns rather than following step-by-step instructions.
He realized that simplifying an equation is like solving a Rubik’s cube: a simple equation can be put together into a complicated equation, and if you know the right moves, you can reverse the process and reduce it to a simpler form.
Shih used this idea to train a machine learning system. First he started with simple equations and deformed them by applying mathematical operations. He then wrote down the steps needed to decipher them. By studying many examples, the AI learned to recognize patterns and reverse the scrambling process.
When the system was given new, complex equations that it had never seen before, it was able to simplify them.
“Our new method achieved a near-perfect simplification rate, far surpassing previous machine learning-based methods,” said Shih.
The research shows that AI could become a powerful tool for symbolic reasoning, the kind of mathematical thinking that scientists use to discover new laws of nature.
Shih said the most unusual part of the project was the way the research was done.
“Claude basically functions here as a graduate student would,” he said. “It did all the practical work that a student would normally do in one of my projects.”
The AI wrote code, ran experiments, generated data, created plots, and helped write the research paper.
The system was incredibly fast and could write and test code faster than a human and could continue to work around the clock. But it made mistakes and sometimes repeated the same mistakes, which meant Shih had to carefully supervise it and check its work.
The experience showed him both the power and limits of artificial intelligence as a research partner.
The project led Shih to a larger question about the future of scientific research and the role AI systems can play.
“Can they achieve total autonomy, or will they simply remain a tool that will make us all much, much more powerful?” Shih said. “I think that’s the trillion-dollar question right now.”
The scientific community is divided between those who say AI could one day make discoveries on its own and those who see a partnership with humans as the guide to AI systems that can work faster and process more data than anyone else, he said.
“There are people who say we’re going to have 10,000 Einsteins, that humans will be obsolete and just watch the AIs do research,” Shih said. “I don’t know if that’s going to happen. I think it’s much more likely that this will allow scientists to do a lot more than they can do now.”
Shih said the shift is already underway in his own lab. He actively trains his postdocs and students to collaborate with AI systems like Claude, teaching them how to supervise and validate the work the system produces. Over time, he said, universities may need to formalize that training through new courses focused on “vibe coding” and “vibe research,” a style of work in which scientists collaborate with AI to explore ideas, test solutions and advance research more quickly.
Shih said he believes AI will become a standard part of scientific research, but human judgment will remain essential.
“The most important skill for the next generation of scientists will not only be solving problems, but also learning to work with AI, mentoring it and validating what it produces,” he said. “If we get this right, the payoff can be enormous in terms of faster progress and new discoveries.”
Discover more ways Rutgers conducts research shapes the future.
