Learning Outcomes
SLO | After this lecture you will be able to... | CLO | Bloom |
|---|---|---|---|
SLO 3.1 | Describe the parts of a Turing machine: tape, head, states and transition rules | CLO1 | C1 Remember |
SLO 3.2 | Trace a simple Turing machine on an input, step by step, and give its output | CLO1 | C2 Understand |
SLO 3.3 | Explain the Church-Turing thesis and the halting problem in simple words | CLO1 | C2 Understand |
SLO 3.4 | Explain the Turing Test, the total Turing Test and CAPTCHA, with their history | CLO1 | C2 Understand |
SLO 3.5 | Explain the Chinese Room argument and the main weaknesses of the Turing Test | CLO1 | C2 Understand |
1. The Turing Machine
§1.1 Why Turing invented it
Slide L03-S06
In 1936 the word 'computer' meant a person who did calculations with paper and pencil by following fixed rules. Alan Turing asked: what exactly can be computed by following simple rules, one step at a time? In his paper "On Computable Numbers" (1936) he described an imaginary machine that does what such a person does. It is a model (a thought experiment), not a real machine - but no real computer can compute anything that a Turing machine cannot; real computers are only faster.
§1.2 The four parts
Slide L03-S07
Part | What it is | Analogy (cashier) |
|---|---|---|
Tape | An unlimited strip of cells; each cell holds one symbol (0, 1, +, or blank ␣) | A very long receipt roll |
Head | Reads the symbol in one cell, writes a symbol, then moves one cell Left (L), Right (R) or Stays (S) | A finger and a pen on one line |
States | The machine's 'state of mind': q0 (start), q1, q2 ... and HALT (stop) | What job the cashier is doing now |
Transition rules | The program: (state, symbol read) → (symbol to write, move, next state) | The rule card |
Read a rule as a sentence: (q0, 0) → (1, R, q0) means "If I am in state q0 and I read 0, I write 1, move one cell right, and stay in state q0." |
2. Tracing Turing Machines
§2.1 Machine 1: the bit flipper
Slide L03-S08 to S09
State | Reads | Writes | Moves | Next state |
|---|---|---|---|---|
q0 | 0 | 1 | R | q0 |
q0 | 1 | 0 | R | q0 |
q0 | ␣ | ␣ | S | HALT |
Trace on input 1011 (head starts on the first symbol). 'Tape before' is the tape when the head arrives at that step.
Step | State | Head at cell | Reads | Tape before | Action |
|---|---|---|---|---|---|
0 | q0 | 0 | 1 | 1011 | write 0, move R, q0 |
1 | q0 | 1 | 0 | 0011 | write 1, move R, q0 |
2 | q0 | 2 | 1 | 0111 | write 0, move R, q0 |
3 | q0 | 3 | 1 | 0101 | write 0, move R, q0 |
4 | q0 | 4 | ␣ | 0100␣ | write ␣, stay, HALT |
Output: 0100. Every bit is flipped.
§2.2 Machine 2: add 1 to a binary number
Slide L03-S10
This copies how you add 1 by hand: go to the rightmost digit; 1 + 1 = 0 with a carry, so keep moving left; when you meet a 0 (or the empty cell in front), write 1 and stop. State q0 walks to the end; state q1 adds the carry.
State | Reads | Writes | Moves | Next | Meaning |
|---|---|---|---|---|---|
q0 | 0 | 0 | R | q0 | walk right over the number |
q0 | 1 | 1 | R | q0 | walk right over the number |
q0 | ␣ | ␣ | L | q1 | end found → step back to the last digit |
q1 | 1 | 0 | L | q1 | 1 + 1 = 0, carry 1 to the left |
q1 | 0 | 1 | S | HALT | 0 + 1 = 1, done |
q1 | ␣ | 1 | S | HALT | carry went past the front → write 1 |
§2.3 Worked trace: 1011 + 1
Slide L03-S11
Step | State | Cell | Reads | Tape before | Action |
|---|---|---|---|---|---|
0 | q0 | 0 | 1 | 1011 | write 1, R, q0 |
1 | q0 | 1 | 0 | 1011 | write 0, R, q0 |
2 | q0 | 2 | 1 | 1011 | write 1, R, q0 |
3 | q0 | 3 | 1 | 1011 | write 1, R, q0 |
4 | q0 | 4 | ␣ | 1011␣ | write ␣, L, q1 |
5 | q1 | 3 | 1 | 1011␣ | write 0, L, q1 |
6 | q1 | 2 | 1 | 1010␣ | write 0, L, q1 |
7 | q1 | 1 | 0 | 1000␣ | write 1, S, HALT |
Output: 1100 (eleven + 1 = twelve). 8 steps (0 to 7).
Second example: on input 111 (seven) the carry runs past the first digit. In state q1 the head reads the blank in front (cell -1) and writes 1, giving 1000 (eight). |
§2.4 How to answer a trace question in the exam
Slide L03-S11
- Copy the rule table. Write the start tape and mark the head (usually on the first symbol, state q0).
- For each step write: step number, state, head cell, symbol read, and the action from the rule table.
- Update the tape after every write. Move the head exactly ONE cell.
- Stop when the state becomes HALT. Write the final tape clearly.
§2.5 Try it in Python
Slide Optional - links to Practical 1
This short program (verified - the output below is its real run) simulates Machine 2. Change the rules and the input tape to test your homework.
# A tiny Turing machine simulator (try it at home in Python) rules = { # (state, read): (write, move, next state) -- Machine 2: add 1 to a binary number ("q0", "0"): ("0", "R", "q0"), ("q0", "1"): ("1", "R", "q0"), ("q0", "_"): ("_", "L", "q1"), ("q1", "1"): ("0", "L", "q1"), ("q1", "0"): ("1", "S", "HALT"), ("q1", "_"): ("1", "S", "HALT"), } tape = dict(enumerate("1011")) # cell number -> symbol state, head, step = "q0", 0, 0 while state != "HALT": read = tape.get(head, "_") write, move, new = rules[(state, read)] print(f"step {step}: {state} reads {read} -> write {write}, move {move}, go {new}") tape[head] = write head += {"R": 1, "L": -1, "S": 0}[move] state, step = new, step + 1 print("final tape:", "".join(tape[i] for i in sorted(tape)).strip("_")) |
step 0: q0 reads 1 -> write 1, move R, go q0 step 1: q0 reads 0 -> write 0, move R, go q0 step 2: q0 reads 1 -> write 1, move R, go q0 step 3: q0 reads 1 -> write 1, move R, go q0 step 4: q0 reads _ -> write _, move L, go q1 step 5: q1 reads 1 -> write 0, move L, go q1 step 6: q1 reads 1 -> write 0, move L, go q1 step 7: q1 reads 0 -> write 1, move S, go HALT final tape: 1100 |
3. What Turing Machines Tell Us
§3.1 Church-Turing thesis and the universal machine
Slide L03-S12
Church-Turing thesis: anything that can be computed by an algorithm (a step-by-step method) can be computed by a Turing machine. It is called a thesis (a strongly believed claim) because it cannot be proved like a theorem. Alonzo Church reached the same idea by a different method in 1936.
Universal Turing machine: one machine that reads the rule table of ANOTHER machine from its tape and runs it. This is the idea of the stored-program computer: your phone is one piece of hardware that runs thousands of apps. Languages such as Python and C++ are Turing complete - they can compute anything a Turing machine can.
§3.2 A limit: the halting problem
Slide L03-S13
The halting problem asks: can one program look at ANY program and its input and say correctly whether it will stop or run forever? Turing proved (1936) that the answer is NO. Idea: if a perfect checker H existed, we could build a program that asks H about itself and then does the opposite of H's answer, so H would be wrong. Problems like this are called undecidable. Everyday link: Windows shows 'Not responding' and asks you to wait or close, because it cannot know whether the program will recover.
4. The Turing Test
§4.1 The imitation game
Slide L03-S15
In "Computing Machinery and Intelligence" (journal Mind, 1950) Turing replaced the vague question "Can machines think?" with a game. A human judge (interrogator, C) types questions to two hidden players: a human (B) and a machine (A). Communication is by text only. After a short conversation the judge must say which is the machine. If the judge cannot reliably tell, the machine passes. Turing predicted that by about the year 2000, after five minutes of questioning, an average judge would have no more than a 70% chance of identifying the machine correctly.
§4.2 Abilities needed and the total Turing Test
Slide L03-S16
Ability | Meaning | Needed for |
|---|---|---|
Natural language processing | Understand and write a human language | Standard and total test |
Knowledge representation | Store what it knows or is told | Standard and total test |
Automated reasoning | Use knowledge to answer and draw conclusions | Standard and total test |
Machine learning | Adapt to new situations and find patterns | Standard and total test |
Computer vision | See and recognise objects (video signal) | Total test only |
Robotics | Handle and move physical objects | Total test only |
§4.3 The Turing Test in practice
Slide L03-S17
Year | Event |
|---|---|
1966 | ELIZA chatbot: simple pattern rules; some users believed it understood them. |
1991-2019 | Loebner Prize: a yearly Turing Test contest; no machine won the gold prize. |
2014 | Eugene Goostman, posing as a 13-year-old Ukrainian boy, fooled 33% of judges in 5-minute chats - criticised as a trick. |
2025 | Researchers at UC San Diego ran a three-party test: GPT-4.5, told to play a shy young person, was judged to be the human 73% of the time. |
§4.4 CAPTCHA: a reverse Turing test
Slide L03-S18
CAPTCHA stands for Completely Automated Public Turing test to tell Computers and Humans Apart (2003). Here the judge is a computer and the one being tested is YOU. Examples: twisted letters, 'select all traffic lights', the 'I'm not a robot' box on bank and admission websites. Its purpose is to block bots that make fake accounts, send spam or buy all the tickets. Modern AI can solve many picture CAPTCHAs, so websites also check mouse movement and behaviour.
5. Can Machines Really Think?
§5.1 The Chinese Room (John Searle, 1980)
Slide L03-S20
A man who knows no Chinese sits in a closed room with a large rule book written in English. Questions in Chinese come in through a slot. He matches the symbols in the book, which tells him which Chinese symbols to send back. People outside receive perfect Chinese answers and think someone inside understands Chinese - but the man understands nothing.
Word | Meaning |
|---|---|
Syntax | The form of symbols and the rules for arranging them |
Semantics | The meaning of the symbols |
Searle's conclusion: a computer only manipulates symbols (syntax) without understanding their meaning (semantics). So passing the Turing Test does not prove understanding. |
§5.2 Replies and weaknesses
Slide L03-S21
- Systems reply: the man alone does not understand, but the whole system (man + book + room) might - just as one brain cell does not understand Urdu but the whole brain does.
- Searle's answer: let the man memorise the whole book; he still does not understand Chinese.
Weakness of the Turing Test | Example |
|---|---|
It tests fooling (deception), not intelligence | Eugene Goostman's excuses as a 13-year-old |
Short tests can be passed with tricks | Five minutes is not enough to find the gaps |
Humans can fail it | A shy person or a non-native English speaker |
It ignores non-human intelligence | A chess engine is very capable but cannot chat |
This is why this course measures AI by rational behaviour (agents, Lecture 4) rather than by fooling people.
6. Common Mistakes
Slide L03-S23
Wrong idea | Correct idea |
|---|---|
A Turing machine is a real, physical computer. | It is an imaginary model of computing (1936). |
The tape is short, like RAM. | The tape is unlimited in both directions. |
The head can jump to any cell. | The head moves only one cell at a time: L, R or S. |
Passing the Turing Test proves a machine thinks. | It shows it can imitate human chat; imitation ≠ understanding. |
CAPTCHA tests computers' intelligence. | In CAPTCHA the computer is the judge and tests whether you are human. |
Word Bank
Word | Simple meaning | Example |
|---|---|---|
Model | A simplified imaginary design used to study something | Turing machine = model of computing |
Transition rule | Instruction: in this state, on this symbol, do this | (q1, 1) → (0, L, q1) |
Halt | Stop running | State HALT |
Algorithm | A step-by-step method to solve a problem | Long addition by hand |
Thesis | A strongly believed claim that is not formally proved | Church-Turing thesis |
Undecidable | No algorithm can ever solve it for all inputs | Halting problem |
Interrogator | The person who asks questions; the judge | Judge C in the imitation game |
Syntax / semantics | Form of symbols / meaning of symbols | Chinese Room |
Unary | Writing a number with only 1s | 3 = 111 |
Summary
- A Turing machine has a tape, a head, states and transition rules: (state, read) → (write, move, next state).
- Trace a machine by recording state, head cell, symbol read and action at every step until HALT.
- Church-Turing thesis: anything computable by an algorithm is computable by a Turing machine. Some problems, like halting, are undecidable.
- Turing Test (1950): a judge chats by text with a hidden human and a machine; CAPTCHA reverses the roles.
- Searle's Chinese Room: symbol manipulation (syntax) is not understanding (semantics).
Self-Check Questions
- Name the four parts of a Turing machine and explain each in one line. (see §1.2)
- Write the rule (q0, 1) → (0, R, q0) as a sentence. (see §1.2)
- Trace the bit flipper on input 0110. (see §2.1)
- Trace Machine 2 on input 111. What is the output and why does the tape get longer? (see §2.2-2.3)
- State the Church-Turing thesis. Why is it called a thesis? (see §3.1)
- What is the halting problem? Give an everyday example. (see §3.2)
- Describe the setup of the imitation game and Turing's prediction. (see §4.1)
- Which two extra abilities does the total Turing Test need? (see §4.2)
- How is CAPTCHA different from the Turing Test? (see §4.4)
- Explain the Chinese Room argument and the systems reply. (see §5.1-5.2)
Reading: Kaplan (2016) Chapter 4; Russell & Norvig (4e) Sections 1.1.1 and 28.1-28.2.