Mike's Notes
Very true.
Resources
- https://sfiscience.substack.com/p/sizing-up-the-ghosts-in-the-machines
- https://sfi-edu.s3.amazonaws.com/sfi-edu/production/uploads/publication/2026/07/31/Parallax-summer-2026-web.pdf
References
- Reference
Repository
- Home > Ajabbi Research > Library > Subscriptions > Santa Fe Institute
- Home > Handbook >
Last Updated
15/09/2026
Sizing Up the Ghosts in the Machines
David Krakauer: David Krakauer, President of the Santa Fe Institute.
...
A BEYOND BORDERS column by David Krakauer, President of the Santa Fe Institute.
...
Diagram of the Human Mind , Robert Fludd, 1617, via Public Domain Review
All of science is in the business of exorcism. When philosopher Gilbert Ryle coined “the ghost in the machine” in The Concept of Mind (1949), he was identifying an unnecessary explanatory entity in Cartesian dualism and expelling it. Descartes had posited a res cogitans, a thinking substance categorically distinct from the mechanical body. Ryle diagnosed this as a “category mistake,” and explained it away in the manner an Oxford professor does any superstition: like being shown the colleges, library, and playing fields of the university, and still asking the benighted question, “But where is the University?”
What Ryle did for mind, other sciences have done for their own spectral inhabitants. But the ghosts differ in kind, and so do their rites of exorcism. Physics exorcises supernumerary particles — unnecessary free parameters in pursuit of universality. Biology exorcises redundant processes — divine or teleological agencies. And the sciences of intelligence exorcise imposed patterns — transcendent sources of cognition. Each tradition seeks minimality, yet each means something profoundly different by it.
The deepest ambition of physics has always been to reduce the number of independent facts the universe requires us to accept. Theoretical physicist Eugene Wigner observed that symmetry and invariance principles act as filters on possible theories, eliminating free parameters with a scandalous efficiency. And Einstein’s general relativity derived gravitation from the minimal geometric postulate that spacetime curvature is gravity, eliminating Newton’s action-at-a-distance.
Biology’s ghost was never a free parameter but an agent. Before Darwin, the complexity of organisms demanded an intelligent designer the way a watch implied a watchmaker. On the Origin of Species (1859) performed an exorcism on the demiurge itself and replaced a transcendent creative process with a mechanical one. The ghost to be exorcised is the unnecessary rule, an act of God or Lamarckian inheritance. What remains after exorcism is not the universal parameterization of physics, but a universal mechanism (selection) that operates locally and iteratively.
Alan Turing showed in 1936 that a device of absurd simplicity consisting of a tape, a head, and a finite set of states could compute anything computable. Warren McCulloch and Walter Pitts demonstrated in 1943 that networks of binary threshold neurons could instantiate logical operations. Both of these exorcisms expelled patterns of architectural extravagance that Descartes and his followers had imposed on the mind. David Marr, in his 1969 model of cerebellar cortex, proposed a single, repeating architecture consisting of Purkinje cells trained by climbing-fiber error signals, that could learn arbitrary input–output mappings. Computation of near arbitrary complication could be solved through scale.
Computer scientists build architectures that seek to be spirit-proof. When physicists criticize biology for having too many parameters, they are applying parametric parsimony where processual parsimony is the goal. And when biologists accuse neuroscience of ignoring evolutionary constraint, they are importing processual parsimony into a domain governed by architectural parsimony. Each field is haunted by a different ghost, exists in a different intellectual bardo, each governed by its own rituals.
The old disciplinary ghosts aspired to be ontological, frightening entities posited to occupy ordered states of reality, from which ensuing and more skeptical generations banished them. We now have epistemic ghosts, patterns in large datasets, manifesting as regularities, correlations, and structures, that no human theorist seems to have previously named or noticed. LLMs suggest that we have been haunted by vast domains of regularity, in language, protein folding, weather, and genomics, that are simply too high-dimensional for human intuition to observe but not too complex for a sufficiently scaled statistical model to discern.
The ghosts are multiplying, many of them hallucinations, and scientific exorcists must learn new rites.
— David Krakauer
President, Santa Fe Institute
From the Summer 2026 edition of the SFI Parallax newsletter. Subscribe for our monthly e-Parallax, or email “news at santafe.edu” to request quarterly home delivery in print.
No comments:
Post a Comment