Deepfakes are spreading through online communities, and people are having a hard time telling the difference between what’s real and what’s not.
If you need a refresher, deepfakes are videos that use machine learning to superimpose famous people’s faces onto other people’s bodies to make it look like they’re doing or saying things that they wouldn’t normally do or say.
According to FastCompany, researchers at SUNY Albany have found a very simple way to help you tell if you’re watching a fake video: blinking.
Healthy adult humans blink every 2-10 seconds, with a single blink taking one-tenth to four-tenths of a second. You would see this if you were watching a normal video of a person talking, but not if you were watching a deepfake video.
The deepfake algorithms are trained on pictures of people’s faces, and there aren’t many pictures of people with their eyes closed — even for the most frequent victims of paparazzi.
Without training images of people blinking, the algorithms have a hard time creating videos where people blink normally. Using machine learning, the SUNY researchers were able to identify deepfakes based on infrequent blinking.
Deepfake technology will inevitably evolve, so this isn’t a permanent solution, but it is a start. Check out Fast Company’s article in The Blurb.
You can take that right to the bank
As blockchain technology gains momentum, the traditional banking industry will need to adapt — or risk being replaced.
The decision to build vs. buy is something we’ll be discussing at the upcoming TRANSFORM Conference.
One of the leaders we’ll be talking about this with is ADP’s chief strategy officer Don Weinstein. Don recently oversaw ADP’s acquisition of WorkMarket — but before the deal, ADP had tried to build the product in-house.
We’ll be reviewing the case study of this acquisition and how Don and ADP ultimately made the call to build vs. buy.