The guide names strong, full, general, weak, narrow and domain-specified AI. Underneath those words there are really three levels, and everything anyone has ever used is the first one.
The three levels, what each could do, and which of them exist — plus the Turing test.
Interactive diagram
Explore the labelled diagram, charts and maps for this topic in full study mode.
Free preview
This is the free notes preview
You're reading the free notes. Aimnova Pro unlocks the full study experience — and you can try it with your first topic free to keep:
- FlashcardsLock in vocabulary and key terms with spaced repetition.
- Practice questionsAnswer exam-style questions and get instant AI marking.
- Mock exams & past-paper vaultSit full mocks and see exactly how examiners award marks.
- Personalised study planA daily plan built around your exam date and weak areas.
| Term | Means | Exists? |
|---|---|---|
| Narrow / weak / domain-specified | Trained for one task | Yes — all of it |
| General / strong / full | Any task a person can do | No |
| Super-intelligent | Beyond human ability at everything | Hypothetical |
Weak does not mean bad: A “weak” AI can beat every human alive at its one task. Weak describes the range of what it can do, not how well it does it — and that is a mark students lose regularly.
Know your predicted grade
Take timed mock exams and get detailed feedback on every answer. See exactly where you're losing marks.
Proposed in 1950: if a person cannot tell a machine from a human in conversation, the machine passes. It is a test of whether the system convinces, not of whether it understands.
Why that distinction matters now
- A system can produce fluent text without any model of what the words refer to.
- Fluency is what people use to judge competence, so a convincing system is trusted further than it has earned.
- This is why “it sounded right” is not evidence, in an exam answer or anywhere else.
Real-world examples you can name
AlphaFold and the protein structure database — AlphaFold 2 published 2021; over 200 million predicted structures released July 2022
A machine-learning system that predicts a protein's three-dimensional shape from its amino-acid sequence, a problem that had resisted fifty years of work. The predictions were released openly rather than licensed.
Who it affected: Biologists worldwide, including laboratories that could never afford the equipment to determine structures experimentally.
The Zelensky surrender deepfake — March 2022
A synthetic video showing Ukraine's president telling troops to lay down their arms circulated on social media and briefly on a hacked news website. It was crude enough to be debunked within hours, but it is the standard example of synthetic media used in wartime.
Who it affected: Ukrainian soldiers and civilians, and anyone judging what video evidence is worth.
How this is tested — you have to name types of AI and avoid arguing from a kind that does not exist. It comes up two ways:
Paper 1 — structured question
- Part a: identify types of artificial intelligence
- Part b: explain why a described system is narrow AI
Paper 2 — source-based question
- Q1: read what a source claims a system can do
- Q2: explain a claim a source makes about AI
The trap: science fiction: Writing about an AI that decides to take over is writing about general AI. Real questions are almost always about narrow systems doing exactly what they were built for.
Identify two types of artificial intelligence.
Model answer plan
See the mark-by-mark plan — for / against / judgement, with marking guidance — in study mode.
Explain one reason why calling a system “weak AI” does not mean it performs badly.
Model answer plan
See the mark-by-mark plan — for / against / judgement, with marking guidance — in study mode.