Artificial intelligence (AI) is already used for many different purposes, and the part it plays in people’s lives is only going to get more prominent. But it can be a difficult topic to get a handle on, both because of its technical nature, and the fundamental philosophical issues it raises.
Recognition, a project that ran from September to November last year, aimed to increase audiences’ understanding of AI by using it to compare British art from the Tate collection with contemporary photojournalism from Reuters.
Created by Fabrica, an Italian communication research centre, Recognition won the IK Prize in 2016, which is given out annually to a project that uses innovative technology to enable the public to enjoy British art from its collection.
The project used AI technology to compare photographs from Reuters’ database (which is updated with new photos in real time) with Tate’s artworks using four different criteria: objects, faces, composition, and context.
Based on these, the system selected images as matches, which were then displayed in an online gallery and at an exhibition at Tate Britain. Visitors to the exhibition could make their own matches and compare them to those of the AI program.
Tony Guillan, who manages the IK Prize at Tate, says that the algorithms made some intriguing matches that encouraged audiences to think critically about how they reacted to images. He points to the example of a photo of a man in a slum in India making utensils to sell, which was matched with an obscure painting of the last supper.
“They had similar colours and composition, eliciting the same emotional reaction, but they were miles apart historically, culturally and geographically. We are very used to critiquing art as a human creation but we forget that photographs are also constructions,” says Guillan.
He says that the judges of the IK Prize wanted to find a project that engaged a broad audience with AI in a conceptual and creative way. “The technology behind Google search, although fascinating and ubiquitous, isn’t going to draw in the crowds or get people thinking about art,” he says.
“Often terms like AI and machine learning aren’t very helpful for non-scientists because they are umbrella terms,” explains Guillan. “I think that Recognition was really good at explaining what AI as a whole is trying to do.”
The Recognition website shows some of the matches that were chosen by the system, together with information about how they were calculated. It’s clear that the system does not think on entirely the same lines as a human being.
A woman in a 1916 painting by Augustus John is incorrectly identified as ‘holding a surfboard’, and in a painting by John Russell exhibited in 1790, the head of a man with long white hair is identified as ‘the head of a dog’.
Guillan says that during the project, the matches produced were “often surprising, and often unintentionally hilarious”. For example, images of Donald Trump, then a presidential candidate, were matched to images of women. But in his view, comparing matches made by the system to those made by humans is a key part of the project’s value. “When you try and make a machine understand images, you inevitably ask ‘how does a human understand images?’,” he says.
Angelo Semeraro, an interaction designer at Fabrica who worked on the project, agrees. “They look like mistakes, but for us, these mistakes are the key to unlock what we are not able to see,” he says. “Maybe AI can allow us to see this match from a perspective that we were not able to before.”
Monica Lanaro, a project manager at Fabrica, says that it was exciting to watch the AI system learning what to do before the project went live.
“It was something non-human that looked alive,” she says. The public’s curiosity about how it worked was also evident, whether they were visiting the exhibition at Tate or interacting with the project online.
“One of the goals was to get people thinking, so that was really something special,” says Lanaro.
Although Recognition has now finished, the project is continuing to generate debate and discussion. Conclusions from a data analysis comparing matches made by humans to those made by the AI system will be published on Tate’s website.
“It will probably have a lot to say about the technology and how advanced or not advanced it was, but it will might also have some nice things to say about humans,” says Guillan.
There are likely to be a number of other museum exhibitions examining the capabilities of AI. The Design Museum’s current New Old pop-up exhibition, running until 19 February, explores how design can enhance people’s later lives.
Objects featured include ElliQ an ‘active ageing companion’ that can suggest digital content or recommend activities in the physical world with the aim of keeping older people active and engaged.
Guillan believes that AI will have a lot to offer museums in future, but says that there are not always firm answers to the issues it raises, especially with regard to art.
“Instead of trying to narrow down the questions, I think you sometimes have to run with the contradictions, and run with the unanswerable,” he says.
Recognition, a project that ran from September to November last year, aimed to increase audiences’ understanding of AI by using it to compare British art from the Tate collection with contemporary photojournalism from Reuters.
Created by Fabrica, an Italian communication research centre, Recognition won the IK Prize in 2016, which is given out annually to a project that uses innovative technology to enable the public to enjoy British art from its collection.
The project used AI technology to compare photographs from Reuters’ database (which is updated with new photos in real time) with Tate’s artworks using four different criteria: objects, faces, composition, and context.
Based on these, the system selected images as matches, which were then displayed in an online gallery and at an exhibition at Tate Britain. Visitors to the exhibition could make their own matches and compare them to those of the AI program.
Tony Guillan, who manages the IK Prize at Tate, says that the algorithms made some intriguing matches that encouraged audiences to think critically about how they reacted to images. He points to the example of a photo of a man in a slum in India making utensils to sell, which was matched with an obscure painting of the last supper.
“They had similar colours and composition, eliciting the same emotional reaction, but they were miles apart historically, culturally and geographically. We are very used to critiquing art as a human creation but we forget that photographs are also constructions,” says Guillan.
He says that the judges of the IK Prize wanted to find a project that engaged a broad audience with AI in a conceptual and creative way. “The technology behind Google search, although fascinating and ubiquitous, isn’t going to draw in the crowds or get people thinking about art,” he says.
“Often terms like AI and machine learning aren’t very helpful for non-scientists because they are umbrella terms,” explains Guillan. “I think that Recognition was really good at explaining what AI as a whole is trying to do.”
The Recognition website shows some of the matches that were chosen by the system, together with information about how they were calculated. It’s clear that the system does not think on entirely the same lines as a human being.
A woman in a 1916 painting by Augustus John is incorrectly identified as ‘holding a surfboard’, and in a painting by John Russell exhibited in 1790, the head of a man with long white hair is identified as ‘the head of a dog’.
Guillan says that during the project, the matches produced were “often surprising, and often unintentionally hilarious”. For example, images of Donald Trump, then a presidential candidate, were matched to images of women. But in his view, comparing matches made by the system to those made by humans is a key part of the project’s value. “When you try and make a machine understand images, you inevitably ask ‘how does a human understand images?’,” he says.
Angelo Semeraro, an interaction designer at Fabrica who worked on the project, agrees. “They look like mistakes, but for us, these mistakes are the key to unlock what we are not able to see,” he says. “Maybe AI can allow us to see this match from a perspective that we were not able to before.”
Monica Lanaro, a project manager at Fabrica, says that it was exciting to watch the AI system learning what to do before the project went live.
“It was something non-human that looked alive,” she says. The public’s curiosity about how it worked was also evident, whether they were visiting the exhibition at Tate or interacting with the project online.
“One of the goals was to get people thinking, so that was really something special,” says Lanaro.
Although Recognition has now finished, the project is continuing to generate debate and discussion. Conclusions from a data analysis comparing matches made by humans to those made by the AI system will be published on Tate’s website.
“It will probably have a lot to say about the technology and how advanced or not advanced it was, but it will might also have some nice things to say about humans,” says Guillan.
There are likely to be a number of other museum exhibitions examining the capabilities of AI. The Design Museum’s current New Old pop-up exhibition, running until 19 February, explores how design can enhance people’s later lives.
Objects featured include ElliQ an ‘active ageing companion’ that can suggest digital content or recommend activities in the physical world with the aim of keeping older people active and engaged.
Guillan believes that AI will have a lot to offer museums in future, but says that there are not always firm answers to the issues it raises, especially with regard to art.
“Instead of trying to narrow down the questions, I think you sometimes have to run with the contradictions, and run with the unanswerable,” he says.