Deepfake technology will add complexity to hiring - Technical.ly

Imagine your top candidate for a crucial leadership position at your organization starts Monday. Tonight you get an anonymous email linking to a video that appears to show that candidate engaged in a hateful public protest.

What do you do next?

Entrepreneurs, HR pros and other hiring managers need to understand that the technology powering deepfakes is developing quickly. Last month, a mother in suburban Philadelphia was accused of using faked videos to incriminate her daughter’s classmates. This will come to the workplace. We’re not well prepared for it.

In 2020, an explosion of free, cheeky mobile apps advanced this work. I’ve been playing with Reface and Avatarify by making goofy photos and videos of my friends and coworkers. It’s all lighthearted at the start.


A certain Technical.ly managing editor’s face on Ron Weasley’s body. (Screenshot via Reface)

Reface is a Silicon Valley startup that raised in December a $5.5 million seed round led by influential VC Andreessen Horowitz. Currently, within 30 seconds, you can match a face from any photo onto any of dozens of preloaded videos, like music videos and movies. Avatarify uses First Order Motion, an open-source component of Snapchat filters, to match your face to another, like a celebrity’s. Last year, Square acquired Dessa, which had used machine learning to allow a speaker to use another cataloged voice — like Joe Rogan.

Neither would yet likely hold up under close scrutiny. These are free-to-download consumer applications using low-quality media and processed in seconds. The singularity is not here yet. Imagine, though, how advanced this technology will get. Artificial intelligence thrills and spooks us exactly because with widespread consumer use, the data can power remarkable gains.

Some think fears are overblown. Twenty years ago I used Adobe Photoshop to make realistic spoofs of my friends. That didn’t transform society in a flash. One CEO who has worked extensively with machine learning told me this week: “In the 90s, the historical scenes in the movie ‘Forrest Gump’ changed public perception of what was possible. That’ll just keep happening.” Slow and steady. Whether you think society will adapt to malleable video or not, you can bet we’ll be in for an uncomfortable transition in the workplace.

Literacy in the limitations (or lack thereof) of video is a good first step for culture builders to understand. Today, background checks for employees are already complex: When do social media messages and photos of a candidate’s personal life cross a line and say something about character? This will be made more difficult when we’ll also have to consider the validity of what we’re seeing.

An HR pro friend of mine told me the first lesson from the rise of deepfakes is that hiring managers will have to learn to pause. If you get that video, you simply can’t rush to remove the offer. Ask the candidate. Seek counsel. Like always, you’ll have to consider whether outside information you receive about a candidate is relevant to their employment. Increasingly you’ll also have to consider whether it’s faked. Seeing won’t always mean believing.