Can AI Be Creative, or Is It Remixing? A Fair Look at Both Sides

[GameChanger]

This week gave us two very different stories about AI and creativity. Nikon disqualified the winner of its Small World in Motion contest after finding that generative AI had been used to create parts of the video. Meanwhile, a musician was sentenced for a fraud that used hundreds of thousands of AI generated songs, played by bots, to collect millions in royalties. Neither story is about AI being creative in a good way. But they raise a question that is worth exploring calmly: can AI be creative at all, or does it only remix what humans have already made?

It helps to start with a definition. The philosopher and cognitive scientist Margaret Boden described three kinds of creativity. Combinational creativity puts familiar ideas together in new ways. Exploratory creativity pushes the boundaries of an existing style or set of rules. Transformational creativity changes the rules themselves, creating something that could not have been imagined within the old framework. This framework is useful because it lets us ask which kinds AI can do, rather than arguing about creativity as a single yes or no question

By that measure, AI is clearly capable of combinational creativity. Ask a model for a poem about quantum computing in the style of a sea shanty and it will produce something new that combines familiar elements. Image models do the same with styles and subjects. Some people call this mere remixing, but a great deal of human creativity works this way too. Many celebrated works of art, music and literature draw heavily on what came before

Exploratory creativity is also within reach. The most famous example came in 2016, when DeepMind's AlphaGo played its now legendary move 37 against Lee Sedol. Professional players initially thought it was a mistake, because it broke with centuries of conventional wisdom. It turned out to be brilliant, and it has since changed how top players think about the game. That was a machine exploring the space of possible moves and finding something humans had overlooked

Transformational creativity is the hardest case. It is not clear that any AI system has yet created a new genre, a new artistic movement or a new way of seeing the world. Part of the reason may be that such shifts usually come from people responding to their lives, cultures and experiences. AI systems learn from patterns in existing data, and they do not have lives or experiences of their own in the human sense. That may change, or it may remain a meaningful difference

There is also the question of intention and meaning. When a person paints a picture or writes a song, they are usually trying to express something. Audiences often value art because of the human story behind it. An AI can produce a technically impressive image, but many people feel something is missing if nobody meant anything by it. Others argue that meaning lives in the audience as much as the creator, and that a moving piece of music moves us regardless of who made it

The law is grappling with this too. In the United States, courts and the Copyright Office have held that works generated by AI without meaningful human authorship cannot be copyrighted. Work that combines human creativity with AI tools can still be protected, depending on how much the human contributed. That is a sensible middle ground in my view. It treats AI as a tool, rather than an author, which matches how most creative professionals use it

There are plenty of positive examples of AI supporting human creativity. The Beatles' final song, Now and Then, released in 2023, used machine learning to separate John Lennon's voice from an old demo recording, something that had not been possible before. The song went on to win a Grammy. Musicians use AI to explore arrangements, designers use it to sketch ideas quickly and writers use it to brainstorm. In those cases, AI expands what people can do rather than replacing them

The problems come when AI is used to deceive. The Nikon case was about rules and honesty, because the competition banned AI generated content and the entry did not meet that standard. The streaming fraud was about using AI to fake both the music and the listeners. In both cases, the issue was not creativity but trust. Clear labelling and honest disclosure would solve much of the tension, letting audiences decide what they value

My own view is that AI can be creative in the combinational and exploratory senses, and sometimes impressively so. Whether it can be creative in the deepest, transformational sense is still an open question. But I do not think we need to settle that to use it well. The most exciting future is one where AI gives more people the tools to make things, while human intention, taste and honesty remain at the centre of creative work

Education is one area where this question has a very practical side. Students now have tools that can produce essays, poems and artwork in seconds. Teachers worry about losing the struggle that helps young people find their own voice. At the same time, AI can show students possibilities they had not imagined and help them get started when facing a blank page. The best classrooms will probably teach both creative skills and thoughtful use of AI, rather than banning one or relying entirely on the other

It is also worth remembering how often new tools have been met with fear in creative fields. When photography arrived, some painters declared that art was finished. When synthesisers appeared, some musicians called them soulless. Both tools went on to create entire new art forms and careers. AI may follow a similar path, opening up creative possibilities for people who never had the training or resources before

Do you think AI can be creative, or is it always remixing? Which of Boden's three kinds of creativity do you think AI can reach? And have you used AI in your own creative projects, whether writing, music, art or something else?

Contest-Winning Microscope Video Thrown Out for Using Generative AIgizmodo.com