• Modern AI owes its existence to insights derived from the physics of complex materials, specifically spin glasses.
• John Hopfield's application of spin glass physics to neural networks in 1982 reinvigorated AI research and enabled machines to learn and recall memories.
• The principles of spin glass physics may be instrumental in enabling AI to imagine and in designing neural networks that are more understandable.

This article explores the unexpected link between spin glass physics and the rise of artificial intelligence. Spin glasses, peculiar metallic materials with puzzling magnetic behaviors, captivated physicists in the mid-20th century. Though they had no practical applications, the theories developed to explain their behavior sparked the AI revolution. John Hopfield, in 1982, used the physics of spin glasses to construct neural networks capable of learning and recalling memories. This conceptual breakthrough allowed researchers to apply physics tools to AI, paving the way for deep learning architectures like ChatGPT.

• What: The application of spin glass physics to neural networks, enabling machines to learn, recall memories, and potentially imagine.
• Who: Key individuals include John Hopfield, Geoffrey Hinton, David Sherrington, Scott Kirkpatrick, Lenka Zdeborová, Dmitry Krotov and Elise Cutts (author). Key organizations include Caltech, IBM Research, Quanta Magazine, Swiss Federal Institute of Technology Lausanne and Bocconi University.
• When: Key events occurred from the mid-20th century to the present, with significant milestones in 1982 (Hopfield's neural networks), 2012 (success of deep neural networks), and 2024 (Nobel Prize for Hopfield and Hinton).
• Where: The research and developments occurred across various institutions and locations, including Caltech, IBM Research, and universities in Milan and Lausanne.

The article highlights the unexpected yet profound impact of theoretical physics on the development of artificial intelligence. It demonstrates how seemingly abstract concepts from condensed matter physics, such as spin glasses and the Ising model, provided the foundation for neural networks and deep learning architectures. The success of generative AI models like ChatGPT is a testament to this connection. The analysis also suggests that the relationship between physics and AI is ongoing, with the potential for future breakthroughs in AI design and understanding through the application of statistical physics.

Hopfield made the connection and said, ‘Look, if we can adapt, tune the exchange couplings in a spin glass, maybe we can shape the equilibrium points so that they can become memories.’

Mathematically, one can replace what were the spins or atoms. Other systems can be described using the same toolbox.

I was looking for a PROBLEM, not a problem. How mind emerges from brain is to me the deepest question posed by our humanity. Definitely a PROBLEM.

The article concludes that the physics of spin glasses, initially deemed a useless pursuit, has had a transformative impact on artificial intelligence. The principles derived from studying these materials have not only enabled the creation of neural networks capable of learning and memory but also hold promise for future advancements in AI, including the development of more understandable and creative AI systems. The ongoing research suggests that statistical physics will continue to play a crucial role in shaping the future of machine intelligence.