Physicists are now using artificial intelligence to design and run real experiments, not just crunch numbers, and the technology has already helped find a new superconducting material.
Quick Take
- Physicists already use AI to write code, do math, and design and run actual experiments, according to a October 2026 report.
- AI helped identify a promising new superconducting material by analyzing data from a Japanese research facility.
- Software has also redesigned parts of the LIGO gravitational-wave detector, making it up to 15% more sensitive.
- Experts say AI has not yet discovered a brand-new law of physics on its own, but it is reshaping how research gets done.
AI Moves From Helper To Collaborator In The Lab
Physicists have used computers for decades to crunch numbers and run simulations. That has changed. A report from New Scientist published October 16 says artificial intelligence is now being used “not only to write computer code and do maths, but even to design and run experiments”. That shift marks a new stage, where AI acts less like a calculator and more like a lab partner.
A community paper built from a 2025 National Science Foundation workshop lays out where AI is already proving useful. Scientists use it to speed up hard theoretical calculations, run frontier experiments more smoothly, and dig through massive, one-of-a-kind datasets that would take humans years to sort through by hand. Those three jobs cover much of what modern physics labs do every day.
Real Discoveries Are Already Happening
The gains are not just theoretical. Researchers used AI to study measurement data from the NanoTerasu Synchrotron Light Source in Japan and found a promising new superconducting material, a substance that could carry electricity with zero resistance. The AI system spotted cause-and-effect patterns in the data that would have been extremely hard for people to find on their own.
AI has also improved instruments scientists already trust. Software came up with an unusual new layout for optical parts inside the LIGO gravitational-wave detectors. That redesign could make the detectors 10 to 15% more sensitive at picking up ripples in space caused by colliding black holes. In plasma physics, a model even caught a symmetry in a simulation that human researchers had missed, forcing them to fix their own interpretation of the data.
Where The Technology Still Falls Short
Despite those wins, AI has not yet produced a brand-new law of physics the way Albert Einstein or Isaac Newton once did. One review of the field notes that “although AI has not yet led to new discoveries in physics, it’s becoming a powerful tool across the field,” helping researchers design experiments and sift through results faster than ever. The honest picture is a technology still building toward bigger breakthroughs, not one that has already arrived.
History offers a useful comparison. Major physics breakthroughs, like Wilhelm Röntgen’s discovery of X-rays or Jocelyn Bell Burnell’s detection of pulsar stars, only happened because new instruments revealed things that had been invisible before. AI may be following that same pattern: a tool that expands what scientists can see and test, setting the stage for discoveries that come later rather than causing them outright.
A Field Learning To Use A Powerful New Tool
Some researchers argue the relationship runs in a surprising direction. One academic paper makes the case that “physics provides the constraints, interpretability, and rigour that AI fundamentally lacks,” suggesting physics has as much to teach AI as AI has to offer physics. That two-way exchange, grounded in decades-old math like the Hopfield model that borrowed ideas from magnetic physics to build early neural networks, shows the fields have been intertwined for longer than most people realize.
The 3:1 Wafer Squeeze: How HBM4, the 2048-Bit Base Die & Warpage Physics Broke the Memory Cycle Forever.
1 HBM4 wafer crowds out 3 server DDR5 wafers. With 10-yr Take-or-Pay LTAs to 2035, downcycles are dead. The 5.5x Davis Double Play deconstructed: https://t.co/Pe0i0r3J6L
— Dawson Lee | Silicon & Infra Alpha (@LiDaosen) October 16, 2026
For now, the honest answer is that AI is transforming how physics gets done, not yet what physics knows. It is speeding up calculations, running experiments, and catching patterns humans miss, all real, measurable changes happening in labs right now. Whether it eventually produces the next Einstein-level breakthrough remains an open question, one the field is actively testing with every new experiment it hands over to a machine.
Sources:
newscientist.com, idiap.ch, iop.org, viirj.org, frontiersin.org, cmu.edu
