Analysis: we can learn a lot about AI from viral images of Donald Trump being arrested and a fashionable Pope Francis that fooled us all
86-year-old Pope Francis is an unlikely source of fashion inspiration. But when a photo purporting to show the pontiff sporting a white puffer coat starting doing the rounds on social media, it quickly went viral. It then emerged the image was a fake, generated by artificial intelligence (AI).
"I just thought it was funny to see the Pope in a funny jacket," Pablo Xavier, the creator of the fake image of Pope Francis, told BuzzFeed News. The 31-year-old used AI tool Midjourney and posted the result on a subreddit thread on March 24, where Reddit user Hagoromo-san later commented: "If the Devil wears Prada , does the Pope wear Balenciaga?"
The image was then shared by a user on Twitter after which it spread widely. To the untrained eye the image was plausible. After all, as reported by men's magazine GQ, the Bishop of Rome’s preference for 'athleisure wear’ threw Vatican tailors for a loop when he was elected. This proximity to authenticity and plausibility is what makes AI generated images and text, like that created by the AI-powered Chat GPT, work so well, says Professor Alan Smeaton, founding Director of the Insight SFI Research Centre for Data Analytics.
An AI tool like Midjourney, or those similar to it like Dall-E 2 and Stability AI, essentially mines images — millions or billions of them — stripping them of all their qualities and context clues: their captions, image descriptions and low-level characteristics, like textures and colours, he explains. You then dive into that "bath" of fragments with a sentence. That sentence is a prompt, which generates a brand new, synthesised image that isn't real. And it's not fragments of the pope’s real face or hands, or a real jacket. But with clever, appropriate and accurate prompt engineering you can pull out images which are credible, says Smeaton.
We can't detect AI images, he says. "We appeal or depend upon people's intuition to believe the distinction between what is and what is not plausible. But that's not a reliable [or] sustainable thing to depend upon." Some days before the picture of Pope Francis went viral, fake AI images of Donald Trump being arrested were also circulated on social media. The image of the pope "was in the grey area of kind of plausible and therefore ended up being humorous, as opposed to Trump in handcuffs," adds Smeaton.
We need your consent to load this rte-player contentWe use rte-player to manage extra content that can set cookies on your device and collect data about your activity. Please review their details and accept them to load the content.Manage Preferences
From RTÉ Radio 1's Today with Claire Byrne, Professor Alan Smeaton talks developments in AI chatbots
The pope has also weighed in, though not explicitly in relation to the fake image. During a conference on Monday at the Vatican, CNN reported that the Pope addressed AI technology. According to Vatican News, he said AI and machine learning had "the potential to contribute in a positive way to the future of humanity" but that this potential would only be realised with "a constant and consistent commitment on the part of those developing these technologies to act ethically and responsibly".
Elon Musk and thousands of other tech experts this week called for a "pause" in the training of AI systems more advanced than OpenAI's ChatGPT-4, until shared safety protocols for such designs were developed, implemented and audited by independent experts. "Powerful AI systems should be developed only once we are confident that their effects will be positive and their risks will be manageable," the letter said.
We are faced with a brave new world. In January, US photo agency Getty Images started legal proceedings against Stability AI at the High Court in London, while US stock image company Shutterstock has partnered with OpenAI so users can create stock images using text-to-image generation with the Dall-E 2 tool. This reflects "two different views on the ethical use or misuse or abuse of those images used to train the AI," says Smeaton.
We need your consent to load this rte-player contentWe use rte-player to manage extra content that can set cookies on your device and collect data about your activity. Please review their details and accept them to load the content.Manage Preferences
From RTÉ 2FM's Dave Fanning, Dr Martin Clancy on on the growing role that AI plays in music
It’s going to become increasingly difficult to identify fake images, says Dr Ricardo Castellini da Silva, Postdoctoral Researcher with DCU Institute for Future Media, Democracy and Society (FuJo) and Media Literacy Coordinator for the Irish hub of the European Digital Media Observatory (EDMO). Every day brings a new challenge or a new problem to address, he says.
"We talk about the limits of expression, what can I say, what I cannot say online. With images it's the same." AI images also bring up the question of copyright, to which we have at least one answer for now: in the US, the Copyright Office announced in February 2023 that AI-created images can't be protected by copyright. But these hyperrealistic AI generators also raise questions about ethics, regulations and the limits of what we should be able to do with this technology.
"Should we establish a limit where you cannot create an imagery with named individuals of any kind?" This would mean you couldn't create images of Pope Francis or Donald Trump, for example. "We need to educate people and one of the most important parts of media literacy is actually the civic responsibility that we have as citizens," Castellini da Silva says.
We need your consent to load this rte-player contentWe use rte-player to manage extra content that can set cookies on your device and collect data about your activity. Please review their details and accept them to load the content.Manage Preferences
From RTÉ One's Claire Byrne Live, CNN's Donie O'Sullivan on the growing use of deepfakes
Some people are scared of the world "regulation" because they think of it in terms of censorship, he says. "But we're not discussing censorship. It's just that we have to regulate how this kind of new technology works. We have to regulate digital platforms, we have to regulate media companies, tech companies, and we have to regulate these kinds of software."
Read more: Why do we continually fall for fake stories and false claims?
"We have to not only talk to specialists, but also to the civil society. We have to bring people to the table and say: what kind of world do you want to live in?" Establishing the limits is the hardest part because people will disagree, he says. But "we need to start discussing this more in society, our responsibility as citizens and the ethical limits of what we do when we are online. The digital world is not different from the real world."
So where is all this going? "No idea," says Smeaton. "But fasten your seat belt, it could be a roller coaster."
The views expressed here are those of the author and do not represent or reflect the views of RTÉ
