Machine learning algorithms are capable of performing millions of computations per second, but many of the simplest of human tasks remain incredibly difficult for our machined counterparts. Take putting on a t-shirt as an example: A team of computer scientists from the Georgia Institute of Technology needed to go through hundreds of thousands of trials to develop an A.I. capable of getting an animated character to get dressed.
This isn’t just a weird flex, the researchers hope that these insights could help lead to the design of assistive robots that can help take care of human beings young and old. Indeed, aging demographics are one of the key drivers of robotics innovation, according to a recent industry report, which notes that the global population over sixty will more than double from 962 million last year to more than 2.1 billion in 2050.
These robots will presumably need to understand how a shirt goes from folded laundry pile to discretely covering the body, but there are other applications. Notably, developing the A.I. required incorporating a physics engine into a machine learning technique called deep reinforced learning. This has robots or simulated characters do a certain task an incredible amount of times until they perfect it (this is the same technique Google’s DeepMind used to develop A.I. that can master perfect information board games like chess).
This is #17 on Inverse’s list of the 20 Ways A.I. Became More Human in 2018.
Because of the physics engine, Alex Clegg, one of the Ph.D. students involved in this study told Inverse that this form of A.I. could immediately be implemented by 3D animators to enable more realistic 3D storytelling. But he’s most excited about its future potential.
“It is exciting to imagine the host of problems we can solve with deep reinforced learning,” he says. “We look forward to continuing working toward enabling robotics and finding solutions to big problems which affect the everyday lives of so many people.”
Until then we can just enjoy the comedy stylings of an animated character clumsily getting dressed. Don’t act like you haven’t gotten yourself caught in a button down when you’re in a rush to get to work in the morning.