Dangers of AI And Deepfakes
Deepfake technology uses Artificial Intelligence (AI) or advanced computing to generate convincing images, audio, or even video, that are not real. Right now you can search online for "AI image of (political candidate) (action)" and will likely find a match. With this technology, bad actors can "create" images or events that did not happen.
Deepfakes can be used for a variety of malevolent purposes:
- sway voters
- phishing/theft/social engineering (the "CEO's voice" calls with wire transfer instructions)
- garner "likes" on social media
- bullying/harassment
- revenge/hurt someone's reputation
- manipulate stock prices
- pornography
- disinformation
- identity theft
- fraudulent job applications
- falsifying evidence
- extortion
Deepfakes are also being used in positive ways, such as modeling, stock photos, and in movies to de-age actors or license a likeness after death. Meta already launched a text-to-video service to generate short video ads.
How to Spot Deepfakes
There are some telltale signs now, but unfortunately technology will improve over time.
- unnatural eye movement
- lack of blinking
- unnatural body shape
- unnatural hair
- awkward posture
- jerky head movement/turning
- lip-syncing effect
- skin tone
- weird lighting/shadows
There is an evolving industry of services to detect deepfakes.
How to Counter Deepfakes
Europol recommends avoiding them rather than spending effort to detect them. Use live video or in person meetings. However even Europol points out the effectiveness of live video will decrease over time as computing speed improves sufficiently to generate the video on the fly.
One recommendation is to have the person perform some random action such as wave their hands a certain way or stand up and turn around.
Long Term Effects
Usually, we work from a default of "truth" for things we see and hear. If not sufficiently countered, society may find itself unwilling to trust news reports or other sources, based on an attitude of, "none of it's real."
If AI does sufficiently take over video production, it could feasibly replace Hollywood and music recording. Will it eliminate creativity? No. But if a studio has a choice to pay Taylor Swift $100 million or pay for some computing time to generate an album their AI predicts will be a success, with no royalties, which might they choose?
September 2023
Send this article to a friend!
Subscribe to The ITS Connection
Related articles