TECHNOLOGY
Kolmogorov-Arnold Networks
The alternative neural network architecture—based on the Kolmogorov-Arnold representation theorem—that Tegmark's MIT group developed in 2024 to improve interpretability and scientific accuracy.
Kolmogorov-Arnold Networks (KANs) are a neural network architecture Tegmark's research group at MIT introduced in April 2024, offering an alternative to the standard multi-layer perceptron (MLP) that has dominated deep learning. KANs replace fixed activation functions on nodes with learnable activation functions on edges, drawing their mathematical foundation from the Kolmogorov-Arnold representation theorem—a 1957 result showing that any multivariate continuous function can be represented as a sum of univariate functions. The architectural shift produces networks that are, for certain classes of problems, more accurate and substantially more interpretable than MLPs: the learned edge functions can often be read as identifiable mathematical expressions, making the network's reasoning visible in ways that opaque MLP weights are not.
In The You On AI Field Guide
KANs illustrate both the promise and the difficulty of the interpretability research Tegmark argues is essential to the wisdom race. The architecture represents a genuine advance in the ability to understand what a network is doing and why—a necessary complement to the scaling-based capability gains that dominate industrial