A new form of artificial intelligence (AI) can create its own films by simulating possible outcomes based on a picture.
As reported by Scientific American, humans are able to look at a picture of, say, two other beings interacting, and imagine different ways the scene could play out, and new technology could allow machines to do something similar.
The Massachusetts Institute of Technology (MIT) put this technology to the test, pitting two AI systems against each other. One AI system would create short videos realistic enough that they tricked the other, which was tasked with trying to pick out real videos from machine-created. When the videos were shown to workers at Amazon’s Mechanical Turk, 20 percent of the time they picked the AI created videos over the real ones.
These videos, Scientific American points out, are about one to one and-a-half seconds long with a 64x64 pixel resolution, so the systems won't be creating episodes of television shows or directing movies anytime soon. But, according to researchers, this kind of technology may help "robots and self-driving cars navigate dynamic environments and interact with humans, or let Facebook automatically tag videos with labels describing what is happening."
"Our algorithm can generate a reasonably realistic video of what it thinks the future will look like, which shows that it understands at some level what is happening in the present," explained Carl Vondrick, a Ph.D. student in MIT’s Computer Science and Artificial Intelligence Laboratory.
It's worth reading Scientific American's entire report for more information into the pretty bizarre technology.
For more science news, you can read about NASA's current problem with space poop in space suits.
Blake Hester is a KY-based writer. Follow him on Twitter @metallicaisrad.
Aucun commentaire:
Enregistrer un commentaire