In fact, it’s the basis for a somewhat unique art installation by David Bowen, associate professor of sculpture at University of Minnesota Duluth. Catchily titled flyAI, the project involves a colony of living houseflies whose fate is determined by the accuracy of artificial intelligence software.
It works by using a camera that’s linked to a Raspberry Pi, as well as TensorFlow machine-learning image-recognition software. As flies land by the camera, they have their picture taken in what Bowen described to Digital Trends as a “fly selfie.”
“It’s an image-recognition system designed to rank the likelihood that what it is seeing is a fly,” he said. “If it determines that it is a fly, it turns on a small pump and delivers water and nutrients.” On the other hand, should the algorithm classify the fly as, well, anything other than a fly, the colony does not get food and water.
The AI-augmented ecosystem has been running for the past 30 days and, as project notes explain, it is “set up to run indefinitely with an indeterminate outcome.”
But don’t worry, animal lovers: Bowen says he would intervene were something to go wrong — although it remains an interesting microcosm of our wider reliance on AI systems.
As impressive as deep learning technology is, it’s still not perfect, as illustrated by the fact that Bowen’s flies are frequently mischaracterized as totally different things, including, an article by Motherboard claims, light switches and egg nog. Should such systems be given so much power? Bowen’s work poses the question, but doesn’t answer it.
“I’m not anti-artificial intelligence,” Bowen said, noting that he had been inspired to create the project after reading philosopher Nick Bostrom’s book Superintelligence: Paths, Dangers, Strategies. “I think it has made our lives easier and safer in a lot of ways. I think self-driving cars are going to be a lot safer than regular cars, for example. But I think it’s important to question. This technology is coming and we need to be smart about how it’s implemented.”
After all, if AI systems get as smart as Bostrom thinks they will, tomorrow we might be the metaphorical flies!
- A.I. doesn’t usually forget anything, but Facebook’s new system does. Here’s why
- Facebook’s new image-recognition A.I. is trained on 1 billion Instagram photos
- OpenAI’s GPT-3 algorithm is here, and it’s freakishly good at sounding human
- The high-tech quest to detect COVID-19 via voice
- Clearview AI’s facial-recognition app is a nightmare for stalking victims