Work #1548 · Mid period

Intelligent Machinery

Turing's 1948 unpublished National Physical Laboratory report on machine intelligence

Alan Turing · 1948 · English · Technical report (NPL, unpublished until 1969)

Tradition: Computer science / philosophy of mind / artificial intelligence

Turing's 1948 'Intelligent Machinery' — the earliest detailed AI research programme, including neural-network-like 'unorganised machines'

Composed in 1948 as an in-house report to the National Physical Laboratory (where Turing had moved from Bletchley Park to design and build the ACE — the Automatic Computing Engine, an early stored-program computer), 'Intelligent Machinery' was not published during Turing's lifetime: the NPL's director Sir Charles Darwin (grandson of the biologist) considered it 'a schoolboy essay' and filed it away. It was finally published in 1969 in B. Meltzer and D. Michie (eds.), 'Machine Intelligence 5' (Edinburgh University Press). The report is Turing's earliest detailed proposal for an artificial-intelligence research programme — eighteen months before his much-better-known 'Computing Machinery and Intelligence' (Mind 59, 1950) where the Turing Test was proposed. 'Intelligent Machinery' covers: the abstract definition of machine intelligence; the distinction between learning by experience and learning by instruction; 'unorganised machines' (Turing's primitive form of neural network: a randomly-connected network of binary-threshold elements that could be trained by selective interference, anticipating both Hebbian learning and the modern perceptron); the use of search and game-playing (chess in particular) as testbeds; the role of training versus innate equipment; and the famous suggestion that a machine could be allowed to 'roam the countryside' as a way of acquiring rich experience. Many ideas later associated with the AI of the 1980s and 1990s (neural networks, reinforcement learning, embodied AI) appear here in nascent form. The report is a remarkable document of pre-symbolic AI thought; together with the 1950 Mind paper and the 1947 Lecture to the London Mathematical Society, it constitutes Turing's principal contribution to the philosophy of AI.

Author

Editions cited

  • Intelligent Machinery, NPL Report (1948); first published in B. Meltzer and D. Michie (eds.), Machine Intelligence 5 (Edinburgh University Press, 1969)
  • Reprinted in Alan Turing, Mechanical Intelligence: Collected Works of A. M. Turing, vol. 1, ed. D. C. Ince (Elsevier, 1992)
  • Modern accessible edition: B. Jack Copeland (ed.), The Essential Turing (Oxford, 2004), pp. 411-432
  • Commentary: Andrew Hodges, Alan Turing: The Enigma (1983, ch. 7); Jack Copeland, Turing: Pioneer of the Information Age (Oxford, 2012)

School Embodiments

Logicism · 22%
Philosophy of Mind · 22%
Transhumanism / Posthumanism · 28%
Naturalism · 12%
Structuralism · 9%
Realism · 7%
Analytic Philosophy · 8%
Logicism 22%

Extension of computability-theory into machine-learning territory.

"A machine which is to imitate a brain must be trained." (Intelligent Machinery, §3)

Earliest detailed Turing statement on machine intelligence.

"The human computer is the model for the abstract machine." (Intelligent Machinery, §2)

Founding statement of an AI research programme.

"Unorganised machines may be trained to perform definite tasks." (Intelligent Machinery, §6 — neural-network-like proposal)

Naturalistic continuum between brains and machines.

"The brain is essentially a complicated unorganised machine." (Intelligent Machinery, §3)

Structural account of intelligence as machine-implementable.

"Any sufficiently general structure can support intelligence." (Intelligent Machinery, §4)
Realism 7%

Realism about machine intelligence as a genuine phenomenon.

"Whether machines can think is not a verbal question." (Intelligent Machinery, §1)

Analytic-philosophical tradition.

Internal Tensions

Earliest detailed proposal for an artificial-intelligence research programme — many ideas anticipating later neural-network and reinforcement-learning approaches. Sir Charles Darwin's dismissal of the report as 'a schoolboy essay' has become a standard example of administrative failure-of-imagination in the AI literature; its 1969 belated publication came too late to influence the field's early development.

I. Time

1948 composition; 1969 publication (21 years later). Turing was 36 at composition.

Attributes
Extent: Infinite Ontological Status: Substantival Grain: Discrete Freedom: Deterministic Traversability: Linear Direction: Uni-directional Dimensionality: One

II. Space

National Physical Laboratory (NPL), Teddington — Turing's institutional base 1945-48, between Bletchley Park and the move to Manchester.

Attributes
Extent: Infinite Ontological Status: Substantival Curvature: Flat Dimensionality: Three Locality: Local

III. Matter

Single technical report (~30 pages). Form is technical-discursive rather than tightly mathematical; Turing was writing for an audience of NPL administrators.

Attributes
Extent: Infinite Ontological Status: Substantival Conservation: Conserved Dimensionality: Three Locality: Local

IV. Observer

Middle Turing. The observer is the cryptanalyst-turned-computer-designer working out the philosophical-technical foundations of artificial intelligence eighteen months before he would publish them more famously in Mind.

Attributes
Time Instance: Single Space Instance: Single Knowledge Extent: Mediated Knowledge Retainment: Partial Physicality: Embodied Agency: Active Number: Plural Metaphysical Agency: Impersonal

V. Energy

Programmatic-AI energies. The report is the most concentrated single document of Turing's thinking about AI in the immediate post-war period.

Attributes
Extent: Infinite Ontological Status: Substantival Conservation: Conserved Dispersibility: Irreversible

VI. Information

Single unpublished-then-published report. Its post-1969 reception has been substantial: many of its proposals (especially the 'unorganised machines' anticipating neural networks) were independently re-discovered by AI researchers who had not read Turing.

Attributes
Ontological Status: Substantival Cosmic Conservation: Conserved Personal Conservation: Non-conserved Granularity: Discrete

Personas that cite this work

Alan Turing

Personas with the nearest attribute fingerprint

Historical figures whose own classification on the same six-dimensional grid lands closest to this work's. Computed by attribute-agreement on coordinates both address.

Computed school proximity

The work's attribute fingerprint scored against all schools using the same quiz scorer. Useful as a sanity check on the hand-curated embodiments above.

How Intelligent Machinery resolves each dilemma

34 resolved positions across 4 dimensions, including 6 distinctive where the majority of schools go the other way · 23 unaligned.

Each dimension is sorted so minority positions come first. Mainstream positions are folded into an expandable list.

Time · 9 dilemmas · 5 distinctive

Persistence, the future, and the direction of becoming.

Distinctive · only 9% of schools agree (18/202)
Do you really choose?
If the brain is a physical system and physical systems are governed by laws, then every choice is also a chain of causes — which raises the question of what was really left to choose.
Choice is real within a determined order — agency and determinism aren’t opposites.
On this view, the future is determined and you are genuinely choosing. Those aren't contradictory because the determination runs through you rather than around you: your reasoning, deliberation, and assent are the way the determined outcome gets settled. Choice is what it feels like from …
Roads not taken The future is open and you are a genuine origin of it. (69%) · Choice is structural illusion — every event is fixed by the prior state. (10%) · Even if the universe is undetermined, you are not the chooser. (6%)
Distinctive · only 9% of schools agree (18/202)
Are addicts responsible for their addiction?
Addiction looks from one angle like the textbook case of agency failing — a person doing what they don't, in any meaningful sense, want to do. From another angle it looks like agency at work in hard conditions. Which it is depends on what agency is.
The addict is genuinely responsible within a determined order.
On this view, the addict is acting within a determined order but is genuinely acting — making decisions, endorsing or resisting urges, seeking or refusing help. Responsibility attaches not because some uncaused choice happened, but because the addict is the kind of agent through which …
Roads not taken The addict could have chosen otherwise — that's why recovery is real. (69%) · The addict's behaviour is the outcome of causes; 'responsibility' is a useful fiction, not a metaphysical fact. (10%) · Even if the universe is undetermined, the addict isn't the chooser. (6%)
Distinctive · only 9% of schools agree (18/202)
Should we hold AI systems responsible for what they do?
When an autonomous AI takes an action that harms someone, the question of who or what is responsible — the developer, the operator, the model itself — turns on whether the model is the kind of thing that can be a responsible agent.
The AI can be a genuine agent within determined conditions — and therefore genuinely responsible.
On this view, what makes a being responsible is not indeterminism but the kind of process the being is. An AI that deliberates, considers consequences, can be given reasons, and modifies its behaviour on reflection is doing what responsible agency is, even if its underlying …
Roads not taken An AI without a free will is not the kind of thing that can be responsible. (69%) · An AI's behaviour is fully determined by training and input; 'responsibility' applies if at all to its makers. (10%) · Neither AIs nor anyone else are the locus of free agency; the question is the wrong one. (6%)
Distinctive · only 15% of schools agree (31/202)
Is the universe running out of usable energy?
The heat death of the universe — entropy maxed out, no further work possible — is among the more sobering implications of mainstream physics. Whether it is structurally inescapable depends on what kind of finitude the cosmos has.
Both time and matter are unbounded; 'running out' is misframed.
On this view, the cosmos has neither a temporal horizon nor a material exhaustion point. The framing of running out presupposes bounds that the cosmos doesn't have. Energy gradients perpetuate; new configurations emerge; the categories that make heat-death scary don't apply at the cosmic scale.
Roads not taken Time is unbounded but matter is finite; usable energy can fail without time failing. (47%) · Time both has and lacks bounds depending on the level you ask at; finitude is conventional. (26%) · The cosmos has bounds; heat death is a real horizon. (12%)
Distinctive · only 15% of schools agree (31/202)
Are natural resources fundamentally finite, or only practically so?
Whether we can grow our way out of resource constraints — or whether the cosmos sets limits the economy ultimately must obey — depends on what kind of finitude matter has.
Resources are practically inexhaustible on cosmic scales; terrestrial limits are engineering.
On this view, matter and time are both unbounded at the largest scales. Terrestrial resource limits are real engineering and political constraints but not metaphysical ones; the cosmos can in principle support whatever expansion intelligence is capable of.
Roads not taken Time goes on but matter is bounded; we are eventually constrained even with infinite time. (47%) · The finitude question is level-dependent; resource ethics happens at the level that constrains us. (26%) · Resources are finite in the strict sense; living well requires accepting the limit. (12%)
4 mainstream positions
Matter · 7 dilemmas, all mainstream
Observer · 37 dilemmas, all mainstream
Could causation work backwards? Causation runs one way — the arrow of time is real and structural. 68% Is the asymmetry between memory and anticipation a real feature of time, or just of us? The asymmetry is real because time itself has a real direction. 68% Is the arrow of time a real feature of the cosmos, or only of how we describe it? The arrow is real and structural; the asymmetry isn't an artifact of description. 68% Is environmental damage ever truly permanent? Damage is real and permanent on the relevant timescales. There is no recovery; there is only limitation. 66% Can a civilization recover from collapse? Civilizational complexity is hard to build and easy to lose; recovery is at best partial. 66% Does the second law of thermodynamics mean something morally? Entropy is what time is. The moral weight, if any, is the weight of working against the current. 66% When does a person begin? A person exists from conception — when a new being comes into existence. 54% What is marriage? Marriage has a given form — it’s a kind of thing we recognize, not make. 54% What is our place in nature? Active in a real nature — we cultivate, steward, transform. 48% Should we colonize space? Cultivating worlds beyond Earth is the next form of stewardship. 48% Is genetic engineering of food stewardship or domination? Genetic modification is cultivation by other means. 48% Should we trust expert testimony when we can't verify it? Trust expertise whose conclusions a competent mind can in principle reproduce. 32% Is religious revelation a real source of knowledge? Revelation is evaluable by reason — and not above it. 32% Does an LLM 'know' the things it correctly produces? An LLM can produce correct outputs but not reason to them; useful, not knowing. 32% Are coincidences ever more than coincidence? Are the dead morally present to the living? Are there indivisible units of experience? Can prayer for someone far away affect them? Could a fetal brain organoid in a petri dish be conscious? Could an AI have a mind that matters? Do animals have moral standing comparable to humans? Does environmental harm in another country bind me morally? Does history have a direction or meaning? Does meditation reveal something genuinely timeless? Does prayer change God's mind? How is knowledge of reality produced? If a teleporter copied and destroyed you, would you have survived? Is divine omniscience compatible with human freedom? Is memory stored or reconstructed? Is reality fundamentally digital? Is salvation, liberation, or fulfillment individual or communal? Is the late-stage dementia patient still the person their spouse married? Is truth universal, tradition-bound, situated, or constructed? What happens to "you" when you die? What kind of religious-theological authority does the tradition recognize? What makes someone the same person over time? Who is the moral primary — the individual, the community, the cosmos, the class, or the species?
Information · 4 dilemmas, all mainstream
← #1547 On Computable Numbers, with an Application to the Entscheidungsproblem All Works #1549 The Chemical Basis of Morphogenesis →