The guide names AI in science fiction and philosophy, cybernetics, the AI winters, and the singularity. The winters are the most useful, because they let you write a sceptical counter-argument that is historical rather than just doubtful.
Promise, collapse, promise, collapse — and what changed to make this time different.
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.
| Era | The promise | What went wrong |
|---|---|---|
| 1950s–60s | Machine translation and reasoning within a decade | Language turned out to need world knowledge, not grammar |
| 1970s | — | First AI winter: funding withdrawn |
| 1980s | Expert systems encoding what specialists know | They captured what an expert SAID, not what an expert noticed |
| late 1980s | — | Second winter: the systems were brittle and costly to maintain |
| 2010s | Learning from data instead of rules | Still running — the open question is what its limits are |
The common cause: Both winters came from the same thing: systems that worked in a laboratory met a world with more exceptions in it than anyone had written down.
See how examiners mark answers
Access past paper questions with model answers. Learn exactly what earns marks and what doesn't.
This time is different
- Learning from data handles exceptions nobody listed
- The systems are in daily commercial use, not only in labs
- Three things arrived together: data, cheap parallel chips, better methods
The pattern may repeat
- Capability is being predicted from a short run of improvement
- The hardest problems are the rare cases, as they always were
- Investment has outrun revenue before, twice
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 Bletchley Declaration on AI safety — November 2023
The first international statement on frontier AI risk signed by both the US and China, agreeing that the most capable models pose risks worth coordinating on. It is a declaration of intent, with no enforcement.
Who it affected: Governments, model developers, and anyone hoping governance keeps pace.
How this is tested — the history is evidence for a claim about the future, never a timeline for its own sake. It comes up two ways:
Paper 1 — structured question
- Part a: identify a development in the history of AI
- Part c: evaluate a claim about what AI will do next
Paper 2 — source-based question
- Q2: explain a claim a source makes about AI's progress
- Q3: contrast an optimistic and a sceptical source
The trap: certainty in either direction: The top band wants a position that acknowledges the other. “AI will change everything” and “AI is overhyped” are both capped without the counter.
Explain one reason why earlier expert systems failed where machine learning has succeeded.
Model answer plan
See the mark-by-mark plan — for / against / judgement, with marking guidance — in study mode.
To what extent is the current period of AI development different from earlier ones?
Model answer plan
See the mark-by-mark plan — for / against / judgement, with marking guidance — in study mode.