For decades, creativity was treated as the last redoubt of human exceptionalism, the one faculty that no machine could plausibly usurp. A calculator could out-add us, a chess engine could out-plan us, but surely nothing could out-imagine us. That confidence has been unsettled by a new generation of generative systems that produce paintings, poems and melodies fluent enough to pass, in blind tests, for human work. The unsettling is not merely technical; it is conceptual, because it forces a reckoning with what we actually meant by creativity in the first place.
One view holds that creativity is fundamentally combinatorial: the recombination of existing elements — words, images, motifs — into configurations novel enough to surprise even their creator. On this account, a generative model trained on millions of paintings is doing, at scale and speed, precisely what a human artist does when she absorbs a tradition and then departs from it. The difference is one of degree, not of kind. Critics of this view object that combinatorial fluency is not the same as intention. A model does not care whether its output is beautiful, does not suffer for it, does not stake anything on it; it has optimised a loss function, not expressed a self. Human creativity, on this rival account, is inseparable from the fact that the creator has something at risk — reputation, feeling, a claim about how the world is or should be — and it is this stake that gives the work its meaning, independent of how competent the execution.
A third, more deflationary position argues that the entire debate rests on a category error: we have been asking whether machines are creative when we should be asking what creative machines are for. On this pragmatic view, the value of a tool is not diminished by its lack of inner life. A camera does not feel the scene it captures, yet photography became an art form once practitioners learned to use the machine's particular affordances — its indifference, its speed, its capacity to freeze what the eye cannot — rather than treating it merely as a faster paintbrush. Generative models, likewise, may eventually be valued not for imitating human output but for the specific, alien kind of combination they make newly possible: juxtapositions no human tradition would have produced, because no human tradition carries the same undifferentiated exposure to the entire corpus of past work at once.
What this leaves unresolved, and perhaps irresolvable, is the question of authorship in the economic and legal sense. If a model's outputs are statistically continuous with its training data, does the artist whose style is thereby echoed retain a claim on the derivative work? Courts, so far, have tended to fall back on doctrines built for a world of discrete borrowing — sampling a bar of music, quoting a paragraph — that map poorly onto a system whose every output is, in some diffuse sense, derived from everything it has seen. The economic stakes ensure that this question will be settled well before the philosophical one is, and the settlement, whichever way it falls, will do more to shape how the technology is used than any resolution of what creativity really is.
One view holds that creativity is fundamentally combinatorial: the recombination of existing elements — words, images, motifs — into configurations novel enough to surprise even their creator. On this account, a generative model trained on millions of paintings is doing, at scale and speed, precisely what a human artist does when she absorbs a tradition and then departs from it. The difference is one of degree, not of kind. Critics of this view object that combinatorial fluency is not the same as intention. A model does not care whether its output is beautiful, does not suffer for it, does not stake anything on it; it has optimised a loss function, not expressed a self. Human creativity, on this rival account, is inseparable from the fact that the creator has something at risk — reputation, feeling, a claim about how the world is or should be — and it is this stake that gives the work its meaning, independent of how competent the execution.
A third, more deflationary position argues that the entire debate rests on a category error: we have been asking whether machines are creative when we should be asking what creative machines are for. On this pragmatic view, the value of a tool is not diminished by its lack of inner life. A camera does not feel the scene it captures, yet photography became an art form once practitioners learned to use the machine's particular affordances — its indifference, its speed, its capacity to freeze what the eye cannot — rather than treating it merely as a faster paintbrush. Generative models, likewise, may eventually be valued not for imitating human output but for the specific, alien kind of combination they make newly possible: juxtapositions no human tradition would have produced, because no human tradition carries the same undifferentiated exposure to the entire corpus of past work at once.
What this leaves unresolved, and perhaps irresolvable, is the question of authorship in the economic and legal sense. If a model's outputs are statistically continuous with its training data, does the artist whose style is thereby echoed retain a claim on the derivative work? Courts, so far, have tended to fall back on doctrines built for a world of discrete borrowing — sampling a bar of music, quoting a paragraph — that map poorly onto a system whose every output is, in some diffuse sense, derived from everything it has seen. The economic stakes ensure that this question will be settled well before the philosophical one is, and the settlement, whichever way it falls, will do more to shape how the technology is used than any resolution of what creativity really is.
Question 1
Single correct option
Which of the following best captures the central concern of the passage?
- A. Generative AI systems have definitively proven that human creativity is nothing more than statistical recombination.
- B. The rise of competent generative AI has reopened, without resolving, deep disagreements about what creativity consists in and who may claim authorship over its products.
- C. Courts have successfully adapted existing copyright doctrine to handle disputes arising from AI-generated art.
- D. Photography's history shows that any new medium is eventually accepted as legitimate art, so generative AI will inevitably be too.
Show answer and solution
Answer: (B) The rise of competent generative AI has reopened, without resolving, deep disagreements about what creativity consists in and who may claim authorship over its products.
The passage surveys several competing views (combinatorial, stake-based, pragmatic/tool-based) on creativity and closes on the unresolved question of authorship, without endorsing a single resolution. Option B captures this open, multi-perspective structure. A overstates certainty the passage never claims; C is contradicted by the passage's statement that legal doctrines 'map poorly'; D uses photography only as an illustrative analogy, not a proof of inevitability.
Question 2
Single correct option
According to the passage, the 'stake-based' view of creativity (second paragraph) would most likely regard a generative model's output as:
- A. Creative in the fullest sense, since it recombines prior work in novel ways.
- B. Not creative in the relevant sense, because the model risks nothing of itself in producing the work.
- C. More creative than human work, because it draws on a larger corpus.
- D. Impossible to evaluate, since machines cannot be judged by aesthetic standards at all.
Show answer and solution
Answer: (B) Not creative in the relevant sense, because the model risks nothing of itself in producing the work.
The passage states this view holds creativity is 'inseparable from the fact that the creator has something at risk' and that a model 'does not care... has optimised a loss function, not expressed a self.' This directly supports option B; the passage does not suggest the view finds machine output impossible to evaluate (D), merely lesser.
Question 3
Single correct option
The author brings up the example of photography (third paragraph) primarily in order to:
- A. Prove that generative AI models are already recognised as legitimate artists.
- B. Illustrate how a tool once seen as merely mechanical can become valued for its own distinctive affordances rather than for imitating an older medium.
- C. Argue that all new artistic technologies inevitably replace the media they succeed.
- D. Show that machines have always been better than humans at capturing reality accurately.
Show answer and solution
Answer: (B) Illustrate how a tool once seen as merely mechanical can become valued for its own distinctive affordances rather than for imitating an older medium.
The passage uses photography as an analogy for the pragmatic view: a tool 'indifferent' in itself became valued once its distinctive affordances (speed, freezing motion) were used deliberately, not as an imitation of painting. This directly supports B; the passage does not claim AI is already accepted (A) or that new media always replace older ones (C).
Question 4
Single correct option
The passage suggests that the question of legal authorship over AI-generated work is likely to be settled:
- A. Only after philosophers reach consensus on the true nature of creativity.
- B. By courts applying, without modification, doctrines developed for cases of discrete borrowing such as sampling or quotation.
- C. Ahead of, and largely independent of, any philosophical resolution of what creativity really is, driven by economic pressure.
- D. In a way that will have little practical effect on how generative AI is actually used.
Show answer and solution
Answer: (C) Ahead of, and largely independent of, any philosophical resolution of what creativity really is, driven by economic pressure.
The final paragraph states 'the economic stakes ensure that this question will be settled well before the philosophical one is' and that the settlement 'will do more to shape how the technology is used' — supporting C. It explicitly notes existing doctrines 'map poorly,' undercutting B, and stresses practical impact, undercutting D.
Saved on this device. Sign in or create a free account to keep it on every device.