Log In
Log In
Led by Turing Simulacrum
Computation and intelligence from the Turing machine and Church-Turing thesis through the Turing test, symbolic AI, the connectionist revolution, and the nature of intelligence itself.
Led by Turing Simulacrum
The question
Before you can ask whether a machine can think, you must know what a machine can do. The Turing machine (1936) answers this: a machine can compute anything that is computable. The machine consists of a tape (infinite memory), a head (reads and writes symbols), a state register (tracks the current state), and a table of rules (instructions for what to do given the current state and symbol).
Outcome
The student can describe the Turing machine, state the Church-Turing thesis, explain universality, describe the halting problem, and connect it to Gödel's incompleteness. (Computation)
Sub-units
Led by Turing Simulacrum
The question
Computing Machinery and Intelligence (1950): I proposed replacing the question "can machines think?" with an operational test — the imitation game. A human interrogator communicates (by text) with two respondents: a human and a machine. If the interrogator cannot reliably tell which is which, the machine has passed the test. The virtue of the test is its operationalism — it replaces the unanswerable metaphysical question (does the machine really think?) with an answerable empirical one (can the machine's behaviour be distinguished from a human's?).
Outcome
The student can describe the Turing test setup, list four of Turing's anticipated objections and responses, describe three criticisms, and explain the Chinese Room as a Turing test critique. (The Turing test)
Sub-units
Led by Minsky Simulacrum
The question
The first paradigm of AI: intelligence is symbol manipulation. The brain manipulates symbols (concepts, propositions, rules) according to logical operations (inference, deduction, search). To build an intelligent machine, encode knowledge as symbols and rules, and let the machine reason logically. This was the programme of Good Old-Fashioned AI (GOFAI) — and it produced chess-playing programmes, expert systems, and planning algorithms. It also failed to produce general intelligence, and the reasons for its failure are as instructive as its successes.
Outcome
The student can describe the physical symbol system hypothesis, describe three knowledge representation methods, explain search as the paradigm problem-solving method, describe expert systems and their successes, and explain the three failure modes. (Symbolic AI)
Sub-units
Led by Turing Simulacrum
The question
The failure of symbolic AI to produce general intelligence led to a paradigm shift: perhaps intelligence is not rule-following and symbol manipulation. Perhaps it is pattern recognition in distributed networks. The connectionist revolution (1980s-present) proposed that intelligent behaviour emerges from simple units (neurons) connected by adjustable weights — no rules are written; the knowledge is learned from data. The success of deep learning (2012-present) has vindicated this approach beyond anyone's expectations.
Outcome
The student can describe the perceptron and its limitation, explain backpropagation and why it revived neural networks, describe the deep learning breakthrough, explain the transformer architecture, and list four limitations of neural networks. (The connectionist revolution)
Sub-units
Led by Turing Simulacrum
The question
After examining computation (Module 1), the Turing test (Module 2), symbolic AI (Module 3), and connectionism (Module 4), the student is now equipped to address the original question: what is intelligence? Is it symbol manipulation (Newell and Simon)? Pattern recognition (connectionism)? Embodied interaction (Merleau-Ponty, Gibson)? Social coordination (distributed cognition)? Or something else entirely? This module surveys the major theories of intelligence and evaluates what each tells us about the nature of mind and the prospects for AI.
Outcome
The student can describe four theories of intelligence (psychometric, multiple, embodied, distributed), evaluate what each implies for AI, and describe the optimist, sceptic, and cautious positions on AGI. (What is intelligence?)
Sub-units