David Wolpert

David Wolpert is a professor at the Santa Fe Institute whose work spans machine learning, information theory, and the physics of computation. He completed a physics Ph.D. at UC Santa Barbara in 1989 after studying neural networks and generalization. At SFI, he published “Stacked generalization” in 1992 and, with William Macready, published the no-free-lunch results for optimization in 1997. At NASA Ames, he studied distributed control with interacting reinforcement-learning algorithms; in 2018, he formalized limits on what inference devices inside a universe can predict, remember, or observe. In December 2025, he published a framework for universe simulation; as of October 2026, he is an SFI professor and works with Jan Korbel on identifying computation in natural systems.

Now at: Santa Fe Institute and Arizona State University
Works: 443
Citations: 32,546
h-index: 48
Badges: Bridge
Wayfarer
Loyal

Publication numbers from OpenAlex, as of 19 October 2026.

Timeline

1989

Neural networks and physics

He completes a physics Ph.D. at UC Santa Barbara in 1989 after studying neural networks and how best to generalize from their inferences.
1991–1996

Stacked generalization

He works as a postdoctoral fellow at SFI from 1991 to 1996. In 1992, he publishes “Stacked generalization,” a method for combining learning algorithms.
1997

No-free-lunch results

In 1997, he and William Macready publish “No Free Lunch Theorems for Optimization.” The result says search and optimization algorithms perform equally on average across all problems in the class.
1998–2010

Collective intelligence at NASA

In 1998, he, Kevin Wheeler, and Kagan Tumer study collective-intelligence systems of interacting reinforcement-learning algorithms that maximize global utility in distributed-control problems. In 2010, he returns to Los Alamos as a Ulam Distinguished Scholar in the Center for Nonlinear Studies.
2013

Joining SFI resident faculty

In September 2013, he joins SFI's resident faculty on a half-time basis and is also a member of Los Alamos National Laboratory's Information Sciences Group.
2018

Limits on scientific knowledge

In 2018, he formalizes inference devices and proves that each has limits on what it can predict, remember, and observe within the universe it inhabits.
December 2025

Formalizing universe simulation

In December 2025, he publishes a mathematical framework for what it means for one universe to simulate another.
As of October 2026

Professor at the Santa Fe Institute

As of October 2026, he is a professor at the Santa Fe Institute. His work spans information theory, machine-learning limits, and the thermodynamic costs of computation. In 2026, he and Jan Korbel describe criteria for identifying what computations arbitrary dynamical systems perform.