Oct 2, 2020 - Technology

Get ready for a flood of deepfakes, experts warn

facial recognition AI

Illustration: Sarah Grillo/Axios

If social media platforms don't start dealing much more aggressively with altered audio and video, they risk seeing their platforms devolve into a sea of faked content, experts tell Axios.

Why it matters: The platforms are already struggling to deal with manipulated media, and the technology to create "deepfakes," which are fabricated media generated by machine-learning-based software, is improving rapidly.

The big picture: Experts have been sounding the alarm about the coming era of deepfakes for some time, but platforms have yet to craft a clear and unified game plan for thwarting them, beyond some internal policy updates. The threat comes as Facebook, Twitter, YouTube, TikTok and other platforms are already under fire for spreading misinformation globally around vital topics including COVID-19, elections and ethnic strife.

  • "We have to stop the line of, 'I don't want to be the arbiter of truth,'" UC Berkeley professor and deep fake expert Hany Farid said during an Axios virtual event. "It is nonsense." Farid notes that all those platforms already draw lines, such as prohibiting pornography, because they see allowing such content as bad for business.

Between the lines: The platforms have been slow to figure out which manipulated media should be taken down, which labeled and which left alone.

  • But the challenges will only grow over the next couple of years as the tools to create faked video and audio become both more powerful and easier to use, and as platforms hesitate to take drastic action against misinformation in the face of GOP claims that anti-conservative bias is built into Silicon Valley's content moderation practices.
  • "I think we've already lost precious time because of the politicization of this issue," said Nina Jankowicz, disinformation fellow at the Wilson Center.

Yes, but: Prohibiting all manipulated media is also tricky, given the everyday practice of editing video for wide distribution and the fact that videos are also often manipulated for satirical purposes.

Be smart: The mere existence of the technology also allows politicians to dismiss real unflattering footage as fake.

What's next: There are several technology solutions aimed at addressing the issue, including an effort to create authenticated content streams to verify that content hasn't been significantly altered from capture through processing and viewing.

Go deeper: The half-hour event is worth a watch if you have time: You'll find it here.

Go deeper