WILL the public eventually reach a point when they no longer trust genuine photographs and videos because artificial intelligence-generated content has become so convincing?
Generative AI Association of Malaysia president David Woo contends that the risk is real and in some ways it’s more dangerous than the fakes themselves; to a certain extent, it may end up as a liar’s dividend.
“When everything can be faked, people start dismissing real evidence as ‘just AI’.
“A politician caught on camera can claim the footage was generated. Researchers call this the liar’s dividend, and it’s a genuine concern.”
And the impact on society is already taking place.
“History explains why. Every era has had to prepare for new possibilities. The railway age, the industrial age, the Internet age, the chip age, each one disrupted how people judged what was real and whom to trust, and each time, society built a new way of trusting.
“Trust didn’t disappear. It was rebuilt on a new foundation. The same will happen here. Once the public can tell the real from the generated, trust will return.
“Think about film. A movie today can be photorealistic in every frame, yet nobody in the cinema believes it actually happened. We’ve all learned to read a film as a film.
“However convincing AI images become, we’ll reach the same point: everyone will know they’re generated, and they will no longer carry the weight of evidence on their own,” says Woo.
According to him, what changes is the mechanism of trust due to the abuse of technology.
“Look at phone calls. Not long ago, everyone answered every call. Today, nobody picks up an unknown number without thinking, and we’ve built a whole new set of instincts around whom to trust on the phone. We didn’t stop using phones. We built a new layer of trust on top of them.
“Photographs are going through the same shift. A photo used to be believed at first glance. Now, a photo of anything important will need verification, and society will adjust.”
Woo points out that in times of uncertainty, people will need a trusted filter to help discern the truth as the raw flow of images can no longer be taken at face value.
“The media will become more important, not less. Credible media that verify, source, and attribute will become the platform that separates real from fake. People will trust an image because of who vouches for it, not because of how it looks.
“Behind that, detection tools running on platforms and in newsrooms will do the heavy lifting the public never sees.
“The goal isn’t a society that distrusts everything. It’s a society with a new, workable way of knowing what’s real. Every era before this one has managed exactly that, and this one will too.”
Consequently, what practical steps can users take to ensure they are not unintentionally spreading misinformation?
“The practical responsibility sits mostly with how we share.
“If your AI content depicts real people or events and could be mistaken for real, label it. Never impersonate a real person for gain.
“Before forwarding anything shocking, take a pause and think: Is the source credible, is anyone else reporting it, does anything look off? If you’re unsure, don’t forward it. Misinformation depends on people sharing before they think.
“And the most important fix is public understanding, so that people read AI images the way they already read cartoons.”
Woo issues a reminder about the responsible use of AI-gene-rated images, their framing, and a general consensus on misuse.
“Creation is creation, whether the medium is watercolour, Photoshop, or AI. We never asked painters to be ‘responsible painters’, or put a warning on every photo touched up in Photoshop.
“Putting that burden on AI creators is a little unfair, and it treats the medium as the problem when the medium is neutral.
“Where we do need consensus is on misuse: using an image to deceive, defraud, or impersonate, or passing something false along as if it were real. That’s about how an image is used, not how it was made.”
He draws an example of how satirical art is different from an AI-generated image which is presented as a genuine event.
“For example, an AI image of a politician in an absurd situation is doing exactly what political cartoons have done for 200 years. Nobody mistakes a cartoon for a photograph.
“The only difference now is that the public hasn’t yet learned to read AI images the same way. The line is crossed when someone presents it as a real event, and existing law on fraud and defamation already covers that.”
