AI researcher predicts flood of fake multimedia content
A Google AI researcher argues that people will soon have to be far more skeptical of videos and images online. New AI algorithms are making it easier than ever to create convincing fakes.
Google AI researcher Ian Goodfellow calls it "a little bit of a fluke, historically" that humanity has been able to trust video evidence for the past few decades. When something was caught on camera, people could generally believe it actually happened.
Goodfellow thinks that era is coming to an end. The reason: easy-to-use and increasingly powerful AI systems called Generative Adversarial Networks, or GANs. These networks pair two neural networks together—one generates content while the other evaluates it. This setup lets the system improve its output on its own, without human oversight.
AI systems now need far less training data
According to Goodfellow, these networks can produce impressive results from relatively little data. Where an AI once needed 60,000 sample images of numbers to learn to read them, today it only needs 100.
When it comes to fake news, Goodfellow says people will need to become much more skeptical. They should stop automatically trusting images and videos they see online. Fakes have always been possible, but AI systems are accelerating the process dramatically.
As an example, Goodfellow points to a recent research project that used GANs to generate countless fake images of supposed celebrities (see cover image and video). The images look authentic, even though the people in the photos never existed. For Goodfellow, this represents genuine creative output from an AI. More information about the project can be found here.
In the past, people managed to stay informed and form opinions without relying on videos and photos. "In this case AI is closing some of the doors that our generation has been used to having open," Goodfellow says.
AI News Without the Hype – Curated by Humans
As a THE DECODER subscriber, you get ad-free reading, our weekly AI newsletter, the exclusive "AI Radar" Frontier Report 6× per year, access to comments, and our complete archive.
Subscribe nowAI news without the hype
Curated by humans.
- Over 20 percent launch discount.
- Read without distractions – no Google ads.
- Access to comments and community discussions.
- Weekly AI newsletter.
- 6 times a year: “AI Radar” – deep dives on key AI topics.
- Up to 25 % off on KI Pro online events.
- Access to our full ten-year archive.
- Get the latest AI news from The Decoder.