| No. | Expected answer | Points | Mark |
|---|---|---|---|
| Part A · Understanding the documents | |||
| 1 | Firemen burn books instead of putting out fires (1). “It was a pleasure to burn” shows that Montag enjoys his job and does not question it at first (1). | 2 points | / 2 |
| 2 | It makes Montag think for the first time; it is the start of his critical thought and of his change. | 1 point | / 1 |
| 3 | That books must contain something precious, ideas worth dying for. | 1 point | / 1 |
| 4 | The chatbot gave a customer wrong information about a fare discount (1); the tribunal ordered Air Canada to pay and refused the idea that the chatbot was responsible for its own words: the company is responsible (1). | 2 points | / 2 |
| Part B · Language and key ideas | |||
| 5 | can; cannot; could (0.5 each). | 1.5 points | / 1.5 |
| 6 | Waymo has run a self-driving taxi service in Phoenix and San Francisco. | 1 point | / 1 |
| 7 | AI that is excellent at one task only (0.5); for example Deep Blue at chess in 1997, AlphaGo at Go in 2016, or ChatGPT for text (1). | 1.5 points | / 1.5 |
| 8 | Filter bubble: algorithms show us only content we already like. Deepfake: a fake video or picture made with AI that looks real. Singularity: the hypothetical moment when AI becomes more intelligent than humans in every field (1 point each). | 2 points | / 2 |
| Criterion | Full marks | Half marks | No marks |
|---|---|---|---|
| Task and examples (2 points) | A clear position, one AI that helps, one AI gone wrong and one idea from Fahrenheit 451. (2) | Position given, only one or two examples. (1) | No position or no examples. (0) |
| Speech techniques (2 points) | Addresses the audience, a rhetorical question and the rule of three. (2) | One technique only. (1) | No speech technique. (0) |
| Organisation (2 points) | Introduction, arguments in order, conclusion, varied linking words (however, that is why). (2) | Some structure, simple linking words. (1) | No visible plan. (0) |
| Language (2 points) | Modals, present perfect and opinion expressions, mostly correct; sequence vocabulary. (2) | Errors sometimes block meaning. (1) | Errors often block meaning. (0) |
| Part | Points | Mark |
|---|---|---|
| Part A · Understanding the documents | 6 points | / 6 |
| Part B · Language and key ideas | 6 points | / 6 |
| Part C · Writing: should AI be limited? | 8 points | / 8 |
| Total | 20 points | / 20 |
Ladies and gentlemen, thank you for welcoming me to Silicon Valley. Here, you build the future. But who will make sure that this future stays human?
I strongly believe that AI is a wonderful tool. It already helps doctors read X-rays and mammograms, and Waymo cars drive people around Phoenix and San Francisco with no one in the driver's seat. However, AI can also make mistakes. In 2024, Air Canada's chatbot gave a customer wrong information, and a tribunal decided that the company, not the machine, was responsible.
More than seventy years ago, Ray Bradbury imagined a society where people stared at giant wall screens and firemen burned books. In Fahrenheit 451, nobody asked questions any more, until a girl asked Montag: “Are you happy?” We must not let screens think for us.
That is why I will defend clear rules, like the European AI Act: rules that protect our data, our jobs and our freedom. Innovation, yes. Blind trust, no. Responsibility, always.