Artificial intelligence is entering a new phase in which it increasingly intersects with the mathematical and physical sciences, creating opportunities for both fields to advance together. According to MIT physicist Jesse Thaler, the relationship between AI and these scientific disciplines should not be viewed as a one-way application of technology but rather as a two-way bridge in which each side drives progress in the other.
The current surge in AI capabilities is rooted in decades of research in mathematics, physics, and related fields. These disciplines provided many of the theoretical foundations, datasets, and complex problems that helped shape modern machine-learning techniques. Recent scientific milestones, including Nobel Prize-recognized work linking physics concepts with AI methods and breakthroughs in AI-enabled protein design, highlight the increasingly intertwined nature of scientific research and artificial intelligence.
To explore this evolving relationship, MIT hosted a workshop bringing together experts from astronomy, chemistry, materials science, mathematics, and physics. Participants found that despite their diverse research areas, scientists face similar challenges when integrating AI into their work. The workshop concluded that coordinated investments in computing infrastructure, shared data resources, and interdisciplinary collaboration could accelerate discoveries in both AI and the sciences.
Thaler emphasizes that science can contribute to AI in several important ways. Researchers can use scientific methods to analyze neural networks and identify the principles governing their behavior, helping to explain how complex AI systems function. Scientific problems can also inspire new algorithms and modeling approaches, while scientific reasoning can guide the development of more reliable and interpretable AI technologies.
A key challenge is training a new generation of interdisciplinary researchers capable of working across both domains. Workshop participants highlighted the need for “centaur scientists,” individuals with expertise in both AI and traditional scientific disciplines. Universities can support this goal through integrated curricula, interdisciplinary graduate programs, and joint faculty appointments.
MIT is already pursuing initiatives that align with this vision, including collaborative research institutes and programs that connect computing with scientific discovery. By fostering stronger connections between AI and the mathematical and physical sciences, researchers hope to accelerate breakthroughs while also improving the transparency, reliability, and understanding of artificial intelligence itself.