Bloopers & News 11/01/24

AI: what to look out for on election day.

Tuesday is election day and I want to make sure that everyone is prepared for the deep fakes that might emerge. We’ve been hearing all election cycle about the dangers of these fakes and international interference, but have you really considered how this could manifest directly after an election?

We might see videos of people stealing ballots, people rioting, people burning down polling locations - and how will we know if they are real? My advice is to remain calm and do NOT have a knee jerk reaction that sends you screaming to share on Facebook.

The reality is that AI is NOT perfect (yet) and you will be able to see tell-tale signs that the image is fake.

• Check out the background. Do some faces look overly blurry or “smeared?”

• How long is the clip? AI video is still in its beginning stages and it is difficult to make longer clips in a short amount of time. If it lasts only 3 seconds, be skeptical.

• Is the light realistic or does it look like a dreamy movie scene?

• Have you seen the event from multiple angles? If it is real news, various outlets will cover it and have their cameras in different locations.

I think (hope) that most of us will be able to spot something that is 100% AI. I am a bit more concerned about hybrid-fakes in which the scene has been filmed using actors and only one face has been replaced.

Check out this ad created for the 2022 mayoral race in Shreveport.

There are still tell-tale signs that it is fake (not just the blink-and-you-will-miss-it disclaimer)

• Did you notice the almost bobble head motion of the fake Mr. Perkins, as if his head is not screwed on tight enough?

• He has almost the same exact flat expression in every shot.

• He doesn’t really blink.

To create this ad the creators found a man of similar size and stature to Mr. Perkins. This was the same technique used for the convincing Deepfake Tom Cruise videos.

I recognize that the Perkins ad was made two years ago, but I don’t feel like even today many people could do better without a lot of time. The worry for me is what if someone has already had a lot of time?

For instance, over the summer someone could have filmed a video with a Kamala Harris-like actress and then plugged in her face to say, “I concede to Donald Trump.” If she wins and the video was only revealed to be a fake the next day, I fear the anger and violence the revelation could spark.

Just the seed of doubt is enough to throw us into turmoil. An article from the Harvard Kennedy Ash Center for Democratic Governance and Innovation talks about the dangers of not just the effect of AI but also “AI panic.”

‘AI panic is itself a democracy problem’ . . . as the actual prevalence of deepfakes is minimal but can be greatly expanded by the media. Therefore, the issue with AI and democracy isn’t just the risk of believing false information but the erosion of trust in authentic content.

So, to reiterate, don’t PANIC. Consult your trusted resources and news, show worrying content to people you trust (but don’t do it on social media!), and if you see something too horrible or too good to be true, trust your instincts.

---

NAILED IT

AI still has trouble depicting any woman who is not 18 or 80. I asked for a 50 year old woman with red hair and glasses who looks shocked and angry.

I have noticed that frequently “older” to AI means a redder nose. So middle-aged women either have rosacea and/or a drinking problem? (Actually, that tracks.)

I asked it to age the woman up to 55, and it seems that the difference between 50 and 55 is starting to shop at Talbots.

I then asked for age 60, thinking it would maybe add a wrinkle or two. How silly of me! The cleavage is back, Talbot’s is gone, and this 60 year old is ready to talk about where she was when we first landed on the moon.

---