Seeing is no longer believing. Photos have been faked and manipulated for almost as long as photography has existed.
Now even reality isn’t required for photos to look authentic – just artificial intelligence responding to a prompt. Even experts sometimes have a hard time determining whether someone is real or not. Can you?
The meteoric rise of artificial intelligence has raised the alarm that the technology used to deceive people is advancing much faster than the technology that can identify the tricks. Technology companies, researchers, photography agencies and news organizations are struggling to catch up and try to set standards for the origin and ownership of content.
The advances are already feeding disinformation and are being used to fuel political division. Authoritarian governments have created seemingly realistic news outlets to advance their political goals. Last month, some people fell for images showing Pope Francis donning a puffy Balenciaga coat and an earthquake that devastated the Pacific Northwest, even though none of these events had happened. The images were created using Midjourney, a popular image generator.
On Tuesday, as former President Donald J. Trump turned himself in to the Manhattan district attorney’s office to face criminal charges, artificial intelligence-generated footage appeared on Reddit showing actor Bill Murray as president in the White House . Another image showing Mr. Trump marching in front of a large crowd with US flags in the background was quickly re-shared on Twitter without the disclosure that came with the original post, noting that it wasn’t actually a photo.
Experts fear that the technology could erode trust in the media, government and society. If every image can be fabricated – and manipulated – how can we believe everything we see?
“The tools will get better, they will get cheaper, and there will come a day when nothing you see on the internet is beyond belief,” said Wasim Khaled, CEO of Blackbird.AI, a company that helps clients combat disinformation.
Artificial intelligence enables virtually anyone to create complex works of art, such as those now on display at New York’s Gagosian Art Gallery, or lifelike images that blur the line between what is real and what is fiction. Add a text description and the technology can produce a related image – no special skills required.
There are often hints that viral images were created by a computer rather than captured in real life: for example, the luxuriously coated pope had glasses that seemed to melt into his cheek and fuzzy fingers. AI art tools also often produce nonsensical text. Here are some examples:
However, rapid advances in technology solve many of those shortcomings. The latest version of Midjourney, released last month, is capable of rendering realistic hands, a feat that, remarkably, had eluded early imaging tools.
Days before Trump turned himself in to face criminal charges in New York City, footage of his “arrest” circulated on social media. They were created by Eliot Higgins, a British journalist and founder of Bellingcat, an open source research organization. He used Midjourney to imagine the former president’s arrest, trial, confinement in an orange jumpsuit, and escape through a sewer. He posted the images to Twitter, clearly marking them as creations. They have since been widely shared.
The images were not intended to fool anyone. Instead, Mr. Higgins wanted to draw attention to the power of the tool – even in its infancy.
A new generation of chatbots
A brave new world. A new breed of chatbots powered by artificial intelligence has sparked a battle to determine whether the technology can turn the internet’s economics upside down, turning current powerhouses into has-beens and creating the industry’s next giants. These are the bots you should know about:
The footage from Midjourney, he said, could be used in facial recognition programs that Bellingcat uses to verify identities, usually of Russians who have committed crimes or other abuses. It is not hard to imagine governments or other nefarious actors fabricating images to harass or discredit their enemies.
At the same time, Mr Higgins said, the tool also struggled to create convincing images with people who are not photographed as much as Mr Trump, such as the new British Prime Minister, Rishi Sunak, or the comedian Harry Hill, “who is probably not so much known outside the UK.
In any case, Midjourney was not amused. It suspended Mr Higgins’ account without explanation after the footage spread. The company did not respond to requests for comment.
The limitations of generative images make them relatively easy to detect by news organizations or others tuned to the risk — at least for now.
Still, image banks, government regulators and a music industry trade group have taken steps to protect their content from unauthorized use, but technology’s powerful ability to mimic and adapt complicates these efforts.
Some AI image generators have even reproduced images – a sickening homage to “Twin Peaks”; Will Smith eats handfuls of pasta – with distorted versions of the watermarks used by companies like Getty Images or Shutterstock.
In February, Getty accused Stability AI of illegally copying more than 12 million Getty photos, along with captions and metadata, to train the software behind the Stable Diffusion tool. In his lawsuit, Getty argued that Stable Diffusion diluted the value of the Getty watermark by including it in images that ranged “from the bizarre to the grotesque”.
Getty said the “brazen robbery and freeriding” occurred “on a staggering scale”. Stability AI did not respond to a request for comment.
Getty’s lawsuit reflects the concern of many individual artists – that AI companies are becoming a threat to competition by copying content they don’t have permission to use.
Trademark violations have also become a concern: Artificially generated images have replicated NBC’s peacock logo, albeit with unintelligible lettering, and show Coca-Cola’s familiar circular logo with extra O’s in the name.
In February, the US Copyright Office weighed in on artificially generated images when it reviewed the case of “Zarya of the Dawn,” an 18-page comic book written by Kristina Kashtanova with art generated by Midjourney. The government administrator decided to copyright the text of the comic book, but not the art.
“Because of the considerable distance between what a user can direct Midjourney to create and the visual material Midjourney actually produces, Midjourney users lack enough control over generated images to be treated as the ‘master mind’ behind them,” the office explains in his decision.
The threat to photographers is outpacing the development of legal protections, says Mickey H. Osterreicher, general counsel for the National Press Photographers Association. Newsrooms will increasingly struggle to authenticate content. Social media users ignore labels that clearly identify images as artificially generated, choosing to believe they are real photos, he said.
Generative AI could also make producing fake videos easier. A video appeared online this week in which Nina Schick, an author and generative AI expert, appeared to show how the technology “created a world where shadows are mistaken for the real thing.” Ms. Schick’s face then glistened as the camera pulled back to show a body double in her place.
The video explained that the deepfake was created, with Ms. Schick’s permission, by Dutch company Revel.ai and Truepic, a California company exploring broader digital content verification.
The companies describe their video, which features a stamp indicating it was computer-generated, as the “first digitally transparent deepfake.” The data is cryptographically sealed in the file; tampering with the image breaks the digital signature and prevents the credentials from appearing when using trusted software.
The companies hope that the badge, which is a fee for commercial customers, will be adopted by other content creators to create a standard of trust regarding AI images.
“The magnitude of this problem will accelerate so quickly that it will drive consumer education very quickly,” said Jeff McGregor, CEO of Truepic.
Truepic is part of the Coalition for Content Provenance and Authenticity, a project created through an alliance with companies such as Adobe, Intel and Microsoft to better trace the origins of digital media. The chipmaker Nvidia said last month it was working with Getty to help train “responsible” AI models using Getty’s licensed content, with royalties paid to artists.
On the same day, Adobe unveiled its own image generation product, Firefly, which will be trained using only images that are licensed or proprietary or no longer under copyright. Dana Rao, the company’s chief trust officer, said on its website that the tool would automatically add content references — “like an imaging nutrition label” — that identified how an image was created. Adobe said it also plans to compensate contributors.
Last month, model Chrissy Teigen wrote on Twitter that she had been tricked by the pope’s puffy jacket, adding that “there’s no way I’m surviving the future of technology”.
Last week, one series of new AI images showed the pope, back in his usual robes, enjoying a large glass of beer. The hands appeared mostly normal, except for the wedding ring on the pope’s ring finger.
Additional production by Jeanne Noonan DelMundo, Aaron Krolik and Michael Andre.