The 2024 Nobel Prizes in Physics and Chemistry highlight the pivotal role of neural networks in scientific advancement. John Hopfield’s foundational work is deeply rooted in statistical physics, tracing back to the Lenz-Ising model of ferromagnetism (1925), with subsequent developments of Ising model dynamics by R. Glauber, models of spin glasses and non-ergodic systems. All this contributed to complex systems theory and the emergence of self-organizing associative memory systems, as explored by S. Amari (1972) and Hopfield (1982, 1984), who connected these concepts to statistical physics. Despite their simplicity, these systems form multiple local states, exhibiting non-ergodic computational irreducibility. In "A New Kind of Science" (2002) S. Wolfram argued that there is no shortcut to understand the behavior of complex systems.

Geoffrey Hinton, a cognitive scientist, pioneered the methods to learn internal representations of information in complex networks, contributing to backpropagation algorithms (1986) and deep learning advancements (2015). This trajectory has spurred remarkable progress in machine learning, including the advent of physics-informed machine learning (PIML). From these theoretical foundations, great advancements in artificial intelligence, exemplified by AlphaGo’s triumph over world champions in Go by 2017, have emerged. This has been achieved by AlphaGo system developed by DeepMind, founded by Demis Hassabis, a computational neuroscientist. His idea was to combine insights from systems neuroscience, machine learning, and computing hardware to "solve intelligence" and apply it to a variety of complex challenges. It led to significant breakthroughs, such as the AlphaFold series of programs that effectively addressed a 50-year challenge in biophysics by predicting 3D protein structures with high accuracy from their 1D amino acid sequences.

The implications for science are profound: we now possess tools to tackle complex systems that are computationally irreducible. Recent developments indicate that large multimodal models can demonstrate creativity in generating novel solutions to scientific problems. Furthermore, the evolution of multi-agent systems points toward the potential for universal general intelligence. This year's Nobel Prizes should prompt us to reflect on the transformative wave of machine learning methods shaping our scientific landscape.