In the late 1940s and early 1950s John von Neumann set out to answer a question that sounds biological but is really logical: can a machine build a copy of itself, and can it do so without the copies getting simpler each generation? He never finished the manuscript, and it was completed and edited by Arthur W. Burks and published in 1966 as “Theory of Self-Reproducing Automata” by the University of Illinois Press. Burks also produced technical reports reconstructing the work, including the description used here as the primary source.
Von Neumann’s key construction was a universal constructor living inside a two-dimensional grid of cells, where each cell follows simple rules based on its neighbors. This was one of the first cellular automata. The constructor reads a tape of instructions describing a machine and then builds that machine in the grid. To reproduce, it builds a copy of itself and then copies its own instruction tape into the copy. The design needed cells with a couple dozen possible states and showed that genuine self-replication, including the copying of the description, is logically possible.
The work is striking because von Neumann identified the need to both build from a description and copy that description separately, mirroring the later discovery of how DNA is both translated and replicated. For artificial intelligence and computing more broadly, it founded the study of cellular automata and demonstrated that complex, lifelike behavior, including reproduction, can emerge from many cells running identical simple rules.
For a general reader, the result is a foundational proof that self-building, self-copying systems are not science fiction but a precise logical possibility, an idea that echoes through artificial life, nanotechnology, and emergent computation.