Street cleaners – and our environment – are getting help from an unexpected source: a bunch of very clever birds.
Many people don’t give birds enough credit for being intelligent beings. But they are – crows, for instance, can recognize human faces. Magpies, relatives of the crow, are known for their smartness and trainability, as well as their affinity for picking up shiny objects and finagling with things like locks.
Artificial intelligence researcher Hans Forsberg took an interest in magpies after noticing these clever creatures around his neighborhood. He started focusing his attention on the birds and was able to train them to trade in pieces of litter for food.
Using a combination of mechanical design, electronics, software, and a 3D printer, he created a machine called BirdBox that senses when a bird drops a bottle cap into it and then dispenses peanuts in return.
While the concept was simple enough, creating the food dispenser part took years of labor and testing.
“I spent remarkably much time creating the actual food dispenser,” Forsberg said.
In the end, he came up with a vibrating design which is extended by funnels. Once a magpie deposits the bottle cap into the hole, the reward is triggered to be delivered through a tube. All this is made possible by a Raspberry Pi system, which has a camera monitoring detection system situated right beneath the trash hole to make it capable of responding to a deposit.
Watch the BirdBox in action:
But how did Forsberg actually train train the magpies? “It all ended up quite simple,” he said. “I arranged a ‘scene’ where the bird ‘accidentally’ pushed down bottle caps into a larger funnel, and while they were investigating the BirdBox trying to understand why it randomly distributed food. As the bottle caps fell down the funnel, the metal-detector tracked them and started the machine…then a few more steps (after that).”
Here’s an in-depth explanation by Forsberg of how exactly he trained the magpies:
Wow, what a great idea. Good for the birds and good for the environment. Congratulations, man!