RobotRepublic — If you think fake news and conspiracy theories on the Internet are bad now, look out.
As University of Washington researchers demonstrated week, artificial intelligence software can produce realistic-looking videos of people saying and doing things they never said or did.
As part of their new study [PDF], the researchers used existing audio and video clips of former Pres. Barack Obama to create a mashup that looks astoundingly real.
Check it out below.
The researchers created the Obama doppelganger demo to demonstrate how such techniques can help virtual reality and augmented reality designers add realistic models of real people to their programs. Other possible applications for the technology may be tools and algorithms that dramatically improve videoconferencing.
Earlier this year, the same researchers showed how just analyzing photos and videos on the Internet is the first step toward creating talking, real-looking models of just about anyone — because just about everyone has photos and videos of themselves online.
The process isn’t child’s play yet. To create the Obama video, the researchers said they had to employ a neural net, which scrutinized millions of frames of video of him, capturing millions of details about how his mouth moves, eyes crinkle and tiny facial muscles tighten and release as he speaks.
Fake video creation “will be possible soon,” say researchers
Previously, it was possible to do what the researchers accomplished here, but only with many hours of painstaking, time-sucking work. However, this new technique employs learning algorithm that help the software to learn a great deal automatically from video it find online.
The researchers didn’t try to make their Obama dopelganger say anything he hadn’t said before — they didn’t put words in his mouth. But that doesn’t mean someone else won’t, said Supasorn Suwajanakorn, the study’s lead researcher. Fake video creation are “likely (to be) possible soon,” he said.
There is a silver lining, though.The computer scientists behind this work say the techniques also will inform the creation of tools that can spot fake video.
The University of Washington researchers will deliver the results of their study, which was co-funded by Samsung, Google, Facebook and Intel, at SIGGRAPH on Aug. 2.
Read the study in full and in place, below
For RobotRepublic, I’m Gina Smith.