Fake Images Created by AI Flood Social Media
Pope's Fashion and Trump's Arrest Also Spread
Serious Concerns Over 'Fake News Deluge'
Elon Musk, CEO of Tesla, is reportedly dating General Motors (GM) Chairman and CEO, a famous American politician, and singer Taylor Swift one after another. The Pope is seen strolling down the street wearing a white padded jacket. Could these be real photos?
All of these are fake images created using the latest version of artificial intelligence (AI). The difference is that unlike past AI, which left somewhat 'awkward' parts such as fingers or lighting, the newly updated AI's performance is almost indistinguishable from reality. As the development of generative AI technology, triggered by ChatGPT, is dazzling, concerns are growing about its increasing risks.
Fake image related to Elon Musk (left) CEO created using artificial intelligence. On the right is Mary Barra, Chairwoman of GM. [Image source=AllYourTech Twitter]
The American tech-focused YouTube channel 'AllyourTech' recently released a large number of photos of CEO Musk on the internet. The first photo shows CEO Musk having dinner with Mary Barra, GM Chairwoman.
Afterward, the channel released a second image. This time, it was a photo of CEO Musk holding hands and enjoying a date with Alexandria Ocasio-Cortez (AOC), a Democratic Congresswoman famous as a 'progressive star' in the U.S. Before the shock subsided, a third image was released, showing CEO Musk spending a cozy time with Taylor Swift.
Another fake image revealed by AllYourTech. Unlike the previous AI-generated images, the depiction of fingers is much more natural. [Image source=AllYourTech]
All these photos were created using the latest updated versions of image-generating AIs such as 'Midjourney' and 'Stable Diffusion.' The difference from before is that now they are almost indistinguishable from real photos. AllyourTech warned, "If we can create images of this level on a home computer, imagine what media companies or government agencies could do to you."
'AI that creates photo-like images' existed before. In January, a foreign netizen created fake images that looked like they were taken in the 1990s using Midjourney, which became a hot topic. However, until then, AI showed shortcomings in fields requiring detailed depiction.
This is due to AI's characteristic of excelling at generating the overall picture but being weak in precise details. Therefore, past AI-generated images consistently showed features such as ▲distorted human bodies, ▲abnormal arm lengths, ▲more than five fingers, and ▲excessive blur, and by recognizing these points, one could quickly determine if an image was fake.
However, the latest version of AI has almost overcome these previous limitations. Human body proportions have become perfect, and fingers are depicted much more realistically compared to before. AllyourTech predicted, "(At the current pace of development) soon, finger problems will no longer appear."
Generative AI has much greater impact than existing image synthesis technologies like Photoshop or Deepfake because it can create photos that never existed before. There are concerns that the internet could become a hotbed of fake news due to AI.
A photo of the Pope wearing a white padded jacket was also revealed to be a fake image created by AI. [Image source=Twitter]
For example, recently, a photo of Pope Francis going out wearing a white long padded jacket and a cross necklace attracted attention from netizens. Questions about "What brand is the Pope's padded jacket?" flooded social media. However, it turned out that this image was a fake created by AI.
The controversy over generative AI has also erupted in politics. Fake photos of former U.S. President Donald Trump being arrested surrounded by police circulated on social media. Additional images showing Trump seemingly fleeing or wearing an orange prison uniform were also shared. The photos were reportedly created by Elliott Higgins, founder of the digital data analysis group 'Bellingcat,' but some netizens mistakenly believed that former President Trump had actually been arrested, causing a commotion.
Experts warn that the flood of fake images could cause social crises. Hany Farid, a forensic expert and professor at UC Berkeley, said in an interview with The Washington Post on the 26th (local time), "Ordinary people identify AI images as fake by looking at fingers, but once AI starts to handle these details, that becomes useless," and warned, "The generation of photos that look more plausible and can incite social anger could lead to political risks."
Calls for national regulation of AI are also emerging. The Brookings Institution, one of the largest policy think tanks in the U.S., emphasized in an article published on the 21st of last month, "The main risks of current generative AI are commercial use and malicious use," and stressed, "Separate considerations for these two acts and clear government policy intervention are necessary."
© The Asia Business Daily(www.asiae.co.kr). All rights reserved.
