AGI-26 keynote

Neil Gershenfeld

Neil Gershenfeld is Director of the MIT Center for Bits and Atoms, an interdisciplinary initiative the Center describes as exploring the boundary between computer science and physical science.

“How to turn data into things, and things into data.”

— the Center for Bits and Atoms, stating its own question, cba.mit.edu

One boundary, crossed in both directions

Read the archive in order and the subject never really changes: what happens where information meets matter? The answer keeps getting more literal — from the thermodynamics of a bit, to a molecule used as a computer, to a building assembled the way a message is coded.

  1. Physics of computation

    from the 1980s

    The archive opens in experimental physics and dynamical systems, and arrives at a question about computers themselves: they have, in his words, “many orders of magnitude more thermodynamic degrees of freedom than information-bearing ones (bits)” — a gap that closes as devices approach physical limits, at which point hardware and software “must be understood together.”

    Signal entropy and the thermodynamics of computation (1996) ↗
  2. Quantum computing in a beaker

    1997–1998

    With Isaac Chuang, a proposal for doing quantum computation in bulk nuclear magnetic resonance — a room-temperature liquid rather than a single isolated system. The follow-up implements Grover’s search on chloroform molecules and reports “the first complete experimental demonstration of loading an initial state into a quantum computer”.

    Bulk Spin-Resonance Quantum Computation, Science (1997) ↗
  3. Things that think

    Media Lab, 1990s

    Then the boundary moves outward: electric field sensing that lets ordinary objects notice people, instruments modelled from data, and the argument that computing was about to leave the computer. The book of that argument gives the period its name.

    Applying electric field sensing to human-computer interfaces (1995) ↗
  4. Digital materials

    2013

    The pivotal move, and the one that separates this record from 3D printing. Instead of depositing continuous material, build from a discrete set of mass-produced parts that interlock — a lattice you can assemble, disassemble and reuse, reported in Science as “an elastic solid with an extremely large measured modulus for an ultralight material.”

    Reversibly Assembled Cellular Composite Materials, Science (2013) ↗
  5. The robot that assembles it

    2015–2020

    If the material is discrete, the machine that places it can be simple, and can crawl over what it has already built. This line runs through robotic macrofabrication to mechanical metamaterials and walking machines assembled from the same voxel kit as the structure they walk on.

    Discretely assembled mechanical metamaterials, Science Advances (2020) ↗
  6. Machines that make machines

    2017–2022

    The end of that logic is self-replication. A hierarchical model for building a spacecraft out of a small part set, aimed at “exponential space exploration via self-replicating spacecraft (known as Von Neumann probes)”; then a robotic swarm reported as “capable of serial, recursive” construction — robots that assemble more robots.

    Self-replicating hierarchical modular robotic swarms (2022) ↗
  7. Say it and it exists

    2024–2025

    And the newest work joins the two threads a conference like this one cares about. Generative AI produces a 3D mesh from speech; the mesh is not directly buildable, so the system converts it into an assembly of discrete parts a robot can actually place. Language in, object out.

    Speech to Reality (2024) ↗

At a conference called AGI-26

Most of the roster asks what a mind is or how to build one. This archive asks a question that sits underneath that: what can be made, by what, and how cheaply does the ability to make it spread?

The through-line is discretization. Digital communication became reliable when signals were coded into a finite symbol set; this work argues fabrication becomes reliable, reversible and error-correcting for the same reason — when parts come from a discrete set instead of a continuous deposit. Self-replicating swarms are that argument taken to its end.

The 2024–2025 papers are where it meets the rest of the conference: a generative model proposes a shape, and the discrete-assembly stack is what makes the proposal buildable.

Read Speech to Reality ↗

A lab, not a solo record

After 2013 almost everything here is co-authored with the Center for Bits and Atoms — Benjamin Jenett, Amira Abdel-Rahman, Kenneth Cheung, Will Langford, Miana Smith, Nadya Peek, Alfonso Parra Rubio and others recur across the discrete assembly line of work. The earlier quantum computing papers are with Isaac L. Chuang, and the time-series work with Andreas Weigend.

Selected work

15 entries drawn from the archive, titled exactly as the records read, with his position in the author list computed from each record. Everything else — including the early physics, the time-series prediction work, and the books in their several translations — is in the corpus.

  1. 2022
    Self-replicating hierarchical modular robotic swarms
    Communications Engineering last of 5 authors
  2. 2020
    Discretely assembled mechanical metamaterials
    Science Advances last of 6 authors
  3. 2020
    Discretely assembled walking machines
    Journal of Micro-Bio Robotics last of 2 authors
  4. 2015
    Macrofabrication with Digital Materials: Robotic Assembly
    Architectural Design first of 5 authors
  5. 2012
    How to Make Almost Anything
    Foreign Affairs sole author
  6. 2007
    Microfluidic Bubble Logic
    Science last of 2 authors
  7. 1999
    When Things Start to Think
    Foreign Affairs last of 2 authors
  8. 1998
    Experimental Implementation of Fast Quantum Searching
    Physical Review Letters author 2 of 3
  9. 1997
  10. 1996

Background

Now
Professor at MIT and Director of the Center for Bits and Atoms, which describes itself as an interdisciplinary initiative exploring the boundary between computer science and physical science.
In this corpus
Records from 1981 to 2026: early experimental physics and dynamical systems, NMR quantum computing with Isaac L. Chuang, electric field sensing and physics-and-media work, and — from 2013 — the discrete assembly programme that dominates the recent record.
Books in the corpus
When Things Start to Think and The Nature of Mathematical Modeling appear as reviews and translated editions; Designing Reality, with Alan Gershenfeld and Joel Cutcher-Gershenfeld, appears as a 2017 record.

About this archive

This is an independent archive of Neil Gershenfeld’s publication record, built for Society of Minds Aligned and AGI-26. Records span 1981–2026 and are drawn from OpenAlex; preprints, versions of record, book reviews and translated editions are all kept, so the archive is a reading index, not a bibliometric count.

Biography is limited to what his own MIT pages state. The corpus was checked for the name-collision problem that affects common names; every record retained here lists him among its authors.

Not written by, reviewed by, or endorsed by Neil Gershenfeld. If you are Neil and something here is wrong, the feedback control at the bottom of the page reaches us directly — and the claim bar will hand you the site.

Are you Neil Gershenfeld? Or know Neil Gershenfeld? We'd like to hand this room over. →
Join the constellation