Rutgers Physicist Conducts Full Research Project With AI as Collaborator
David Shih, a physicist at Rutgers University, completed a particle physics research project in full collaboration with Claude, an AI system. The work offers a concrete example of how scientific research is changing as AI systems take on hands-on tasks traditionally done by graduate students.
Shih's team developed a machine-learning method to simplify complex equations in particle physics. The research was posted to arXiv, a preprint server widely used by physicists to share new work before peer review.
From Rubik's Cubes to Particle Physics
Shih noticed that simplifying mathematical equations shares logic with solving Rubik's Cubes. Both involve scrambling and unscrambling problems. He used this insight to train an AI system to recognize patterns in complex equations and reduce them to simpler forms.
The method achieved a nearly perfect simplification rate, outperforming previous machine-learning approaches. Simplifying equations helps physicists see patterns more clearly, make more precise predictions, and reduce computing power needed for calculations.
In particle physics, equations describing subatomic particle collisions can contain hundreds of terms. Physicists expect these complex expressions to hide something simpler underneath, based on fundamental symmetries in nature.
AI Handled the Hands-On Work
Claude wrote code, ran experiments, generated data, created plots, and helped write the research paper. Shih supervised the work, checking results and correcting mistakes.
"Claude is actually functioning here like a graduate student would," Shih said. "It did all the hands-on labor that a student would normally be doing in one of my projects."
The AI worked around the clock and wrote code faster than a human could. But it made errors and sometimes repeated the same mistakes, requiring careful oversight.
Universities Need New Training Models
Jack Hughes, chair of Rutgers' Department of Physics and Astronomy, said the work shows how quickly research practices are changing. "There is an urgent need to train our students and postdocs in this new style of research," Hughes said.
Shih is already training his postdocs and students to collaborate with AI systems. He teaches them how to guide the work and validate what the system produces. Over time, universities may need formal courses focused on what he calls "vibe coding" and "vibe research" - styles of working where scientists partner with AI to explore ideas and test solutions.
The key skill for the next generation of scientists will be learning how to work with AI, guide it, and validate its output, Shih said.
The Bigger Question: Partnership or Autonomy?
The scientific community is divided on AI's future role. Some researchers believe AI could eventually make discoveries independently. Others see a partnership model where humans guide AI systems that work faster and handle more data than any individual person.
"Can they reach total autonomy, or will they just remain a tool that will make us all much, much more powerful?" Shih said. "I think that's the trillion-dollar question right now."
Shih is skeptical of claims that AI will make human scientists obsolete. "I think what is much more likely is that it's going to allow scientists to do much more than they can today," he said.
Working with AI changed what he could attempt as a researcher. "If we learn how to use these tools properly, it will allow us to take on more ambitious problems," Shih said. "It changes the scale of what one person can do."
Human judgment remains essential. The payoff from developing this partnership model could be substantial in terms of faster research progress and new discoveries.
Researchers interested in building these skills may benefit from formal training. Claude AI Courses and AI Research Courses can help scientists learn to work effectively with AI systems in their own projects.
Your membership also unlocks:
