If AI reaches and then passes human-level general intelligence, its goals must be set with great care (remember King Midas), and it will have to be more ethical than we are. That raises two problems: telling machines what 'good' is when we disagree ourselves, and accepting moral machines that limit our freedom. The NHS shows these questions are already here.
The multiple choice (14-19) goes through the first six paragraphs, one each. YES / NO / NOT GIVEN (20-23) covers G, H and J. The summary (24-26) is the NHS example in I.
Multiple choice
Questions 14-19
How to attack this block: Each question names a paragraph and asks for its main point. The first says narrow AI's limits are temporary. The second explains why machines will outperform brains. Midas illustrates badly specified goals, and the fourth paragraph sets the challenge of making AI friendlier than us. The fifth says machines will be ethically superior, and the sixth says our own flaws make the job harder.
Question 14
Answer: CParagraph A, last three sentences
What is the writer's main point about AI in paragraph A?
these prodigious accomplishments are all in so-called narrow AI, where machines perform highly specialised tasks. But many experts believe this restriction is very temporary.
Question says → Passage says
unlikely to see limitationsthis restriction is very temporary
the capabilities of AIhuman-level performance on the full range of tasks
Why: Today's AI is limited to narrow, specialised tasks, but that restriction is 'very temporary', and by mid-century machines may handle the full range of human tasks. So the limits on what AI can do won't last: C.
The trap: A is tempting because 'may' sounds uncertain, but the writer gives a fairly confident timescale ('By mid-century'). D misreads 'many experts believe': nothing says experts disagree. B takes the chess example as the point, when it's only an illustration.
Question 15
Answer: AParagraph B
What is the writer doing in paragraph B?
Machines will be free of many of the physical constraints on human intelligence. Our brains run at slow biochemical processing speeds on the power of a light bulb
Question says → Passage says
machines will be able to outperform humansthere's little reason to think it will stop there
explaining whyMachines will be free of many of the physical constraints on human intelligence
Why: B says AI won't stop at human level, and then gives the reason: our brains are slow, low-powered and size-limited, while machines won't have those constraints. That's an explanation of why machines will outperform us: A.
The trap: D picks up on the comparison of eyes with the Webb telescope, but that's showing how far humans are from the limits, not which human abilities are most advanced. B is wrong because the paragraph is about differences, not shared features.
Question 16
Answer: BParagraph C, last two sentences
Why does the writer bring up King Midas?
the more powerful AI becomes, the more important it will be to specify its goals with great care. Folklore is full of tales of people who ask for the wrong thing, with disastrous consequences
Question says → Passage says
poorly defined objectivesask for the wrong thing / didn't really intend this to apply to his breakfast
can go wrongdisastrous consequences
Why: The writer says AI's goals must be specified carefully, then uses Midas as an example: he asked for everything he touched to turn to gold, without meaning his breakfast. A badly worded wish with a disastrous result is B.
The trap: D is the strongest distractor: gold and 'super profitable' are right next to it. But Midas's gold is a disaster, not a financial advantage, and the profit line is a separate point.
Question 17
Answer: DParagraph D
What challenge does the writer describe in paragraph D?
One thing that makes this task difficult is that we are far from reliably human-friendly ourselves… If superintelligent machines don't do a lot better than us, we'll be in deep trouble.
Question says → Passage says
ensuring AI is more human-friendlycreate powerful AI machines that are 'human-friendly'
than we are ourselveswe are far from reliably human-friendly ourselves / do a lot better than us
Why: The task is to build human-friendly AI, and it's hard because we aren't reliably human-friendly ourselves. The machines must 'do a lot better than us'. That's D.
The trap: B (deciding which values) is the 'getting started' problem, but that's in paragraph F. C picks up 'many other creatures', which is an example of our bad behaviour, not the challenge. A is wrong because the paragraph is about making machines better, not people.
Question 18
Answer: CParagraph E, first and third sentences
What does the writer say about the future of AI in paragraph E?
we want the machines to be ethically as well as cognitively superhuman… Luckily they'll be smart enough for the job.
Question says → Passage says
superior to humanssuperhuman / better than us at finding them
in certain respectsethically as well as cognitively
Why: The writer expects machines to be superhuman in two named areas, ethics and thinking, and 'better than us' at finding the moral high ground. Two specific areas = 'certain respects': C.
The trap: A grabs 'For safety's sake', but that's our safety. The paragraph never treats the safety of machines as an issue. B is wrong because the writer sounds sure ('they'll be smart enough'), and D's opposition comes later, in H.
Question 19
Answer: DParagraph F, middle
What is the writer's main argument in paragraph F?
This won't be easy, given that we are tribal creatures and conflicted about the ideals ourselves. We often ignore the suffering of strangers
Question says → Passage says
Human shortcomingswe are tribal creatures and conflicted about the ideals ourselves / ignore the suffering of strangers
more difficultThis won't be easy
Why: We have to tell machines clearly what 'good' is, and that's hard because we're tribal, divided about ideals and often ignore suffering. Our own flaws make the job harder: D.
The trap: A is tempting because the passage talks about our 'dark sides', but that's in D, and F says nothing about abuses of power. B and C aren't mentioned at all: F isn't about how machines learn or how many people can program them.
YES / NO / NOT GIVEN
Questions 20-23
How to attack this block: G supports the first statement: moral machines might stop us favouring our own communities. H gives the NO, because few people would welcome ethical silicon police. The passage says nothing about how many silicon police would be needed. J says it will take a cooperative spirit.
Question 20
Answer: YESParagraph G, last sentence
Could moral machines stop us favouring our own communities?
We might lose our freedom to discriminate in favour of our own communities, for example.
Question says → Passage says
Machines with the ability to make moral decisionsMachines who are better than us at sticking to the moral high ground
may prevent us frommight lose our freedom to
promoting the interests of our communitiesdiscriminate in favour of our own communities
Why: The writer says morally better machines might discourage our lapses, and the example is losing the freedom to favour our own communities. 'Might lose our freedom to' = 'may prevent us from', so YES.
The trap: 'Discriminate' sounds negative while 'promoting the interests' sounds positive, which makes some people answer NO. They describe the same behaviour from different angles.
Question 21
Answer: NOT GIVENParagraph H, second and third sentences
Does the writer say silicon police would need to be numerous to work?
But are we ready for ethical silicon police limiting our options? They might be so good at doing it that we won't notice them
Question says → Passage says
Silicon policeethical silicon police
to be effectiveso good at doing it
Why: The writer mentions silicon police and says they might be very effective, even invisible. Nothing is said about how many would be needed, so it's NOT GIVEN.
The trap: 'So good… we won't notice them' is about effectiveness, which the statement mentions. But it links effectiveness to skill, not to numbers, and that missing link is the point.
Question 22
Answer: NOParagraph H, last sentence
Are many people happy about machines limiting their freedom?
but few of us are likely to welcome such a future
Question says → Passage says
Many people are comfortablefew of us are likely to welcome
their independence being restricted by machinesethical silicon police limiting our options / sacrificing our own autonomy
Why: 'Such a future' means one with silicon police limiting our options. The writer says few people would welcome it, which contradicts 'many people are comfortable'. NO.
The trap: The start of H says losing some freedoms is good progress (child labour, smoking bans). That's the writer's view of particular freedoms, not people feeling comfortable with machines restricting them.
Question 23
Answer: YESParagraph J, last sentence
Will getting machines to act in our interests take everyone cooperating?
It is not yet clear whether this is possible, but if it is, it will require a cooperative spirit, and a willingness to set aside self-interest.
Question says → Passage says
ensure that machines act in our best interestspoint these new thinkers in the right direction, and get them to act well for us
we all need to work togetherit will require a cooperative spirit
Why: 'This' refers to getting machines to act well for us, and the writer says that will require a cooperative spirit. Need = require, and working together = cooperative spirit. YES.
The trap: 'It is not yet clear whether this is possible' can push people to NOT GIVEN. The doubt is about whether it can be done at all, but the writer is clear about what it would need.
Summary completion (phrase list A-F)
Questions 24-26
How to attack this block: Match meaning, not words. 'How resources are used' becomes resources being allocated. Senior doctors become medical practitioners. The 'control they presently enjoy' becomes professional authority.
Question 24
Answer: CParagraph I, second sentence
What does AI already have a small role in allocating in the NHS?
AI already has some input into how resources are used in our National Health Service (NHS)
Question says → Passage says
a limited rolesome input
the way ___ are allocatedhow resources are used
Why: 'Some input into how resources are used' becomes 'a limited role in the way ___ are allocated'. The plural verb 'are' needs a plural noun, and 'available resources' (C) matches.
The trap: B 'specialised tasks' comes from paragraph A and is also plural, but it's about narrow AI, not the health service. You don't allocate tasks in this summary, you allocate resources.
Question 25
Answer: AParagraph I, fourth sentence
Which group would lose some of their present control?
However, we'd be depriving some humans (e.g. senior doctors) of the control they presently enjoy.
Question says → Passage says
certainsome humans
not having their current level ofdepriving… of the control they presently enjoy
Why: The people who'd lose out are 'senior doctors', and the only phrase in the list that means doctors is 'medical practitioners' (A).
The trap: F 'technology experts' might seem like the people affected by AI, but the passage names doctors, not tech staff.
Question 26
Answer: EParagraph I, fourth sentence
What would those doctors lose?
we'd be depriving some humans (e.g. senior doctors) of the control they presently enjoy
Question says → Passage says
their current level ofthey presently enjoy
professional authoritythe control
Why: What senior doctors 'presently enjoy' is control over decisions. Control in your job is professional authority, so E fits.
The trap: B 'specialised tasks' sounds like a doctor's work and often gets picked here, but the passage is about control, not tasks. D 'reduced illness' is a benefit, not something people lose.
C. 'This restriction is very temporary' means AI's capabilities won't stay limited. A tempts, but the writer is fairly confident about the timescale (by mid-century).
D. The problem is that we aren't reliably human-friendly ourselves, so AI must do better than us. B (deciding which values) is the 'getting started' problem in F, not this paragraph.
E (professional authority). Doctors would lose 'the control they presently enjoy'. 'Specialised tasks' is from the first paragraph and is a distractor here.
Words from this passage worth keeping
prodigious(A)
remarkably great or impressive
handicaps(B)
disadvantages that make something harder
coexistence(C)
living alongside each other
fallible(D)
capable of making mistakes
utopian(F)
imagining a perfect but probably unrealistic world
autonomy(G)
the freedom to make your own decisions
lapses(G)
small failures to behave well
far-fetched(I)
hard to believe; unlikely
Sit this test on Langujet, then come back
The full Cambridge 18 Test 2 reading paper: all three passages, 60 minutes, marked the second you finish with a review of every answer. Already done it? Take it again and see if the same traps still get you.
Cambridge IELTS 18, Test 2, Reading Passage 2. The passage and questions belong to Cambridge University Press and are not reproduced here. We quote short extracts only as evidence for our own commentary, and the answers are published for self-marking. Get the book for the full text, or sit the test on Langujet.
