Often, modern artificial intelligence is praised for its growing sophistication, however mostly in greyer terms. If you are one who expects poor outcomes, the AI revolution will automate millions of jobs, remove the barrier between reality and artifice and force mankind to the brink of extinction eventually.
However, it’s easy to forget that right now, most AI is awfully stupid and useful only in narrow domains for which its underlying software has been particularly trained, such as playing an old Chinese board game or translating text.
If you ask your standard recognition bot to do something new, such as analyze and label a photograph using just its acquired knowledge, you will get quite nonsensical results. That’s what’s interesting about ImageNet Roulette, a neat web tool created as part of a continuing art exhibition on the history of image recognition systems.
Artist and researcher Trevor Paglen, who created the exhibit Training Humans with AI researcher Kate Crawford, explains that the point is not to judge AI but to engage with its present form and its sophisticated academic and commercial history, however grotesque it might be.
Crawford said, “When we first started conceptualizing this exhibition over two years ago, we wanted to tell a story about the history of images used to ‘recognize’ humans in computer vision and AI systems. We weren’t interested in either the hyped, marketing version of AI nor the tales of dystopian robot futures. We wanted to engage with the materiality of AI, and to take those everyday images seriously as a part of a rapidly evolving machinic visual culture. That required us to open up the black boxes and look at how these ‘engines of seeing’ currently operate.”
Even if ImageNet Roulette shows the goofier side of it, it’s an admirable pursuit and an interesting project. It’s mainly an object recognition set, however it has a category for “People” which contains thousands of subcategories, each trying to help software do the apparently impossible task of classifying a human being.
And undoubtedly, Image Net Roulette is terrible at it.