© 2025 KSUT Public Radio
NPR News and Music Discovery for the Four Corners
Play Live Radio
Next Up:
0:00
0:00
0:00 0:00
Available On Air Stations

AI's getting better at faking crowds. Here's why that's cause for concern

SCOTT DETROW, HOST:

This Will Smith video tore through the internet recently...

(SOUNDBITE OF ARCHIVED RECORDING)

WILL SMITH: (Rapping) So dry your eyes and then you'll find a way.

DETROW: ...Not for his performance, but for the crowd. Eagle-eyed viewers noticed odd fingers and faces among the fans and suspected AI manipulation. Crowd scenes, like at concerts, rallies and protests, have long tripped up AI systems, but buckle up because the tech is getting better. NPR's Chloe Veltman has more.

CHLOE VELTMAN, BYLINE: Will Smith's team hasn't publicly commented on how the video was made, but San Francisco-based visual artist and researcher Kyt Janae, an expert on AI image creation, says AI was used in places. She dropped by recently to point out where.

KYT JANAE: That woman's real. That reaction's real. That's not real.

VELTMAN: Janae pauses the video when she comes across glitches in the audience footage.

JANAE: We've got these very long fingers sort of melting into this woman's face. And then it seems like there's - maybe her neck is meshing into somebody's hair.

VELTMAN: Janae says these weird effects are happening because crowd scenes present a particular technological challenge for AI image creation tools.

JANAE: You're managing so many intricate details. You have each individual human being in the crowd.

VELTMAN: They're all moving independently and have unique features. But Janae says AI models, such as Google's Veo 3 and OpenAI's Sora 2, are getting pretty good.

JANAE: We're moving into a world where, in a generous time estimate of a year, the lines of reality are going to get really blurry, and verifying what is real and what isn't real is going to almost have to become like a practice.

VELTMAN: And Janae's observation could potentially have serious consequences in a society where images of big, engaged crowds at public events like rock concerts, protests and political rallies have major currency. Thomas Smith is the CEO of Gado Images. The company uses AI to help manage visual archives.

THOMAS SMITH: We want a visual metric, a way to determine whether somebody is succeeding or not. And crowd size is often a good indicator of that.

VELTMAN: A report from consulting firm Capgemini shows nearly three-quarters of images shared on social media in 2023 were generated using AI. With the technology becoming increasingly adept at creating convincing crowd scenes, manipulating visuals has never been easier. With this, Smith says, comes both a creative opportunity and a societal hazard.

T SMITH: AI is a good way to cheat and kind of inflate the size of your crowd.

VELTMAN: He adds there's also a flip side to this phenomenon.

T SMITH: If there's a real image that surfaces and it shows something that's politically inconvenient or damaging, there's also going to be a tendency to say, no, that's an AI fake.

VELTMAN: Like in August 2024, when then-Republican Party nominee Donald Trump spread false claims that Democratic rival Kamala Harris' team used AI to create an image of a big crowd of supporters. Chapman University emerging technologies lecturer Charlie Fink says it's especially easy to dupe people into believing a fake crowd scene is real or a real crowd scene is fake because of the mode of delivery.

CHARLIE FINK: The challenge is that most people are watching content on a small screen. And most people are not terribly critical of what they see and hear. If it looks real, it is real.

OLIVER WANG: The more realistic and believable we can create the results, the more options it gives people for creative expression.

VELTMAN: Oliver Wang is a principal scientist at Google DeepMind. He co-leads the company's image-generation efforts. Wang says a balance needs to be struck between enabling users to create ever more realistic and believable content, including detailed crowd scenes, and mitigating potential harms.

WANG: Misinformation is something that we do take very seriously. So we are stamping all the images that we generate with a visible watermark.

VELTMAN: However, this watermark is tiny and easy to miss. And AI labeling systems, including invisible watermarks, which Google also uses, are still being applied rather unevenly across platforms. There are still no industry-wide standards.

Meanwhile, Will Smith has been having more fun with AI since that controversial concert video came out. He posted a playful follow-up.

(SOUNDBITE OF ARCHIVED RECORDING)

W SMITH: One, two, three, go.

VELTMAN: The camera pans from footage of the singer performing energetically on stage to an audience packed with fist-pumping cats.

Chloe Veltman, NPR News.

(SOUNDBITE OF LOLA YOUNG SONG, "CONCEITED") Transcript provided by NPR, Copyright NPR.

NPR transcripts are created on a rush deadline by an NPR contractor. This text may not be in its final form and may be updated or revised in the future. Accuracy and availability may vary. The authoritative record of NPR’s programming is the audio record.

Chloe Veltman
Chloe Veltman is a correspondent on NPR's Culture Desk.