Top AI researchers race to detect ‘deepfake’ videos: ‘We are outgunned’ | THE WASHINGTON POST


The ability to create synthetic videos outweighs detection methods by ‘100 to 1’, according to a digital forensics professor at UC Berkeley quoted in THE WASHINGTON POST.

The US public is not ready for the possible consequences of deepfakes in the upcoming Presidential elections, the Post reports. AI researchers are ‘racing to defuse an extraordinary political weapon: computer-generated fake videos.’

  • Differences between real and fake may be imperceptible to the human brain within 18 to 24 months. At the same time, the software is becoming more accessible.
  • Some uses can be highly creative, generating new images and videos never before imagined or expressed by humans.
  • Others create lifelike images and videos that can be put to malicious use, such as political leaders appearing to say things that are untrue, or individuals engaged in derogatory activities.

OTHER RECENT DEVELOPMENTS

  • A deepfake of Mark Zuckerberg attracted attention as a piece of art unapproved by the Facebook founder.
  • Americans believe ‘made up news’ is a bigger problem than violent crime, immigration, terrorism, racism or sexism, according to a new study by Pew Research.

OUR TAKE

  • In a world where video (and audio) can be routinely falsified, the public increasingly relies on news outlets to mediate trustworthy information.
  • Detecting deepfakes goes well beyond current competencies in fact-checking. New skills and new technologies will be needed.
  • Look for an ‘truth arms race’ as tools to create deepfakes become more sophisticated, in turn requiring better tools to identify forgeries.

RELATED STORIES

SEE FULL STORY

Top AI researchers race to detect ‘deepfake’ videos: ‘We are outgunned’ THE WASHINGTON POST | June 12, 2019 | by Drew Harwell

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.