The Turing test provides a popular benchmark for replicating human intelligence, but a better measure may be its lesser-known sibling the Lovelace Test, which measures whether a machine can independently create something original.

The assessment takes its name from Ada Lovelace, the celebrated computer scientist who envisioned machines that could create as well as calculate through an engine that "weaves algebraic patterns, just as the Jacquard loom weaves flowers and leaves".

robot ai art creative computer devrimb
© iStock/Devrimb

Professor Geraint Wiggins is one of the computer scientists teaching machines to pass her test.

The Queen Mary University of London (QMUL) professor holds PhDs in both computational linguistics and musical composition from the University of Edinburgh, and combines the two in his research into computational creativity.

Wiggins believes that machines can use creative thinking to find solutions to human problems that ordinary people can't even imagine.

Take a road junction that gets blocked when it rains and delays the lives of citizens. A human or an ordinary machine would make an obvious recommendation, such as building a new road. A creative computer could conceive of a solution that is beyond the minds of humans.

"Computers can come up with suggestions that people would not think about because computers have the capacity to put together and make connections that humans will not necessarily make," Wiggins explains breathlessly at the and& festival in Leuven, Belgium.

Judging creativity

Machines have long been able to reproduce brushstrokes or sounds and reapply artistic techniques, but are only beginning to understand how to create original art.

Wiggins believes that if machines are to produce art that people like, they have to go beyond doing what they're programmed to do and develop a notion of what humans feel.

"In order to make that happen we really need our machines to be reflective, so it's not just that they do stuff, but they reason about the stuff they do," he says.

His method by which they can do this is called computational creativity, a subset of AI that Wiggins and his colleague Simon Colton define as "the philosophy, science and engineering of computational systems which, by taking on particular responsibilities, exhibit behaviours that unbiased observers would deem to be creative".

Wiggins has been working with Marcus Pearce, a lecturer in sounds and music professor at QMUL, to try to teach computers how humans feel.

The duo has developed statistical models that can calculate the probability of musical notes provoking uncertainty, which can be used to predict how people will respond to sounds.

"The systems that we have can identify structure in language and in music without training," says Wiggins.

These systems enable autonomous composers to create new pieces of music by predicting which notes should be played next in a sequence.

Wiggins argues that by generating original ideas these machines are not only creating but imagining.

"These systems really importantly can predict things that they have never previously perceived," he says.

"Then we're talking about what is really quite creative in the sense that it can generate new stuff that no one has ever generated before, just the same as humans do all the time without even registering."

The machine becomes an artist

Perhaps the most famous AI artist was designed by Harold Cohen, who developed a rule-based computer programme called AARON that creates original artistic image.

AARON's artworks have sold for tens of thousands of pounds and been exhibited at the Tate Gallery.

Compare this to the paintings of Picasso that line galleries around the world, and to the art created by a child that will never find a buyer. Whose work is the most creative?

Google engineer Kenric McDowell is a fan of AARON's works.

"The irony of it is that Harold wanted his system to be able to produce paintings after his death, and he recently passed but from my understanding no one can really understand how to use it. So there's a sort of beauty in the embeddedness of Harold Cohens in that machine."

McDowell is head of the Artists and Machine Intelligence programme at Google, where he brings together artists and AI researchers to develop new insights that can inform Google research.

He thinks that creative computers will augment rather than replace human artists, and has benefited from their help in his other life as an electronic music producer.

Cohen may have had a similar thought in mind when he called his art "only small c creative", but he never had the chance to explain exactly what that meant before he died.

An existential threat?

As machines come closer to passing the Lovelace test, some are worried that this could change what it means to be a human.

Wiggins sympathises with their concerns.

"Creativity is one of the things that we feel very very attached to in our lives," he says. "Computers can be intelligent, and if they can be creative too, what's the point of us?"

His answer is that creative machines will both help humans and stimulate to try new things.

"Creativity does not have to be the great grand thing that we put in a gallery. Creativity is the human way of dealing with life, of dealing with the immense complexity of the world that we have. It's a matter of degree, and it's a matter of quality of output."

Studies have shown that if humans believe an artwork is produced by a machine they are unlikely to like it, even if it was in fact created by a human.

McDowell thinks that their bias reveals a limited understanding of creativity.

"That's kind of a framing problem because if you this art is created by a machine, someone made the machine,” he says with a smile. "It's not very fair to that person."