Authors
Kushel HM is a mechanical engineer-turned-journalist, who loves all things football, tennis and films. He was with the news desk at the Hindustan Times, Mumbai, before joining Newschecker.
Ruby leads editorial, operations and initiatives at Newschecker. In her former avatar at New Delhi Television (NDTV), India’s leading national news network, she was a news anchor, supervising producer and senior output editor. Her over a decade-long career encompasses ground-breaking reportage from conflict zones and reporting on terror incidents, election campaigns, and gender issues. Ruby is an Emmy-nominated producer and has handled both local and international assignments, including the coverage of Arab Spring in 2011, the US Presidential elections in 2016, and ground reportage on the Kashmir issue since 2009.
From a fabricated image of Pope Francis wearing a stylish puffy coat to a set of digitally created prison mugshots of former US president Donald Trump, AI made its presence firmly felt in the world of fake news last year. One AI-generated photo, purportedly showing the team that rescued 41 workers who were trapped inside a tunnel in Uttarakhand in November 2023, had even fooled multiple national media outlets into publishing it as a real image.
2024 will continue to see widely accessible AI tools that allow users to clone voices, generate photo-realistic images, and power bot accounts on social media with human-like conversational skills. Although current generative AI technology is not yet perfect, its ease of use makes it ripe for misuse and it is this potential harm from deepfakes that has left governments, lawmakers, and tech giants scrambling to ensure effective safeguards against AI-generated mis- and disinformation in a crucial year for democracy with elections scheduled across the world, including in India and the US.
Newschecker recently came across two cases of AI-generated media, which signify the growing risk of the technology and how it can be used in contrasting scenarios.
Claim 1
Several social media users are circulating a purported image of US President Joe Biden in military uniform at a national security council meeting, set to authorise military action in the Middle East. These images come against the backdrop of news reports stating that Biden has decided how the US will respond to a drone strike that killed three American troops in Jordan over the weekend. The archived version of the tweet can be seen here.
Fact
Newschecker ran a keyword search for “Biden military uniform”, which led us to this New York Post report, dated January 30, 2024, stating that AI-generated images of Joe Biden in military uniform went viral amid Middle East tensions.
“AI-generated photos showing President Biden decked out in military gear in the Situation Room went viral Tuesday — amid Middle East tensions in the wake of the deadly drone attack on US soldiers. The images — on the heels of the Taylor Swift deepfakes scandal — were viewed thousands of times after first being posted on Elon Musk social media platform on Monday,” read the report.
The report led us to this tweet by @luke_brock, an AI educator, who had first shared the images on January 30, 2024. A subsequent tweet acknowledged that he generated the images using AI.
We also ran the images past Hive Moderator, a leading AI-generated content detection tool, which stated that the images were 99.99% likely to be AI generated.
Result: False
Claim 2
On a lighter, yet eerie note, a purported 01:32-second wildlife short documentary, claiming the existence of strange fish that bear an uncanny resemblance to human faces has gone viral. The archived version of the tweet can be seen here.
We received this video on our Whatsapp tipline (9999499044), too, requesting us to fact-check it.
“In the isolated tropical region of Karanji, there is Lake Samsara, one of the lakes of the tropical sources of the Nile. Recently, strange species of fish that resemble human faces have been discovered, which has left aquatic research scientists confused and astonished….. UNBELIEVABLE!!?,” read the claim, which stated that the species was called Homo Piscis.
Fact
Newschecker did not find any instance of a Lake Samsara, while we learnt that there was a Karanji Lake, but in Karnataka, contradicting the viral claim of it being located near the Nile river in Africa.
We then ran a search for “Homo Piscis”, which did not throw up any credible news reports or accounts of such an aquatic species. We saw that the video was uploaded by Headtap Videos, as seen in the closing credits of the viral short film, on Youtube on January 18, 2024. “Experiments in video and graphics,” read the Youtube bio of Headtap Videos, indicating it was digitally created.
We ran screengrabs of the video past Hive Moderator, a leading AI-generated content detection tool, which stated that the images were 98.8%-99.7% likely to be AI generated.
We also came across multiple reports, seen here, here and here, debunking the existence of the Homo Piscis. According to the reports, a comment under the original video read, “Absolutely shockingly untrue! AI is powerful. It’ll fool a lot of the unschooled,” to which the video creator responded, “Good thing you were here to point out the obvious.” We, however, could not find the comment, indicating that it may have been removed.
Result: False
A Quick Guide On How To Spot AI-Generated Images
AI-generated images and videos have their own oddities, like slight mismatches between sound and motion and distorted mouths. They often lack facial expressions or subtle body movements that real people make, according to an NPR article, dated June 13, 2023.
How To Tell AI Images Apart
- Look for a watermark
- Look for disproportions
- Unlikely settings and architecture
- Look for unrealistic hair patterns
- Look for unnatural hand positions and weird fingers
- Look for garbled text
- Watch out for extremely sharp features or overly smooth texture
For example: There are at least two of the irregularities usually associated with AI-generated images seen in the recent Biden case. In one of the images (left), you can see that Biden has six fingers, a telltale sign of AI-generated content, while the text patch on Biden’s uniform (right) is in gibberish.
However, experts also caution against relying too heavily on these kinds of tells. “The newest version of Midjourney, for example, is much better at rendering hands. The absence of blinking used to be a signal a video might be computer-generated, but that is no longer the case,” read the NPR report, adding that one model to detect AI images, created by research scientist Mike Caufield, is called SIFT. That stands for four steps: Stop. Investigate the source. Find better coverage. Trace the original context.
Plus, run it past an AI image detection tool for good measure.
Also Read: Video of Monkeys Ringing A Bell Is Not From Ayodhya Ram Temple
Sources
New York Post report, January 30, 2024
Tweets, @luke_brock, January 30, 2024
Hive Moderator, AI-generated content detection tool
Youtube video, Headtap Videos, January 18, 2024
If you would like us to fact-check a claim, give feedback or lodge a complaint, WhatsApp us at 9999499044 or email us at checkthis@newschecker.in. You can also visit the Contact Us page and fill out the form. Follow our WhatsApp channel for more updates.
Authors
Kushel HM is a mechanical engineer-turned-journalist, who loves all things football, tennis and films. He was with the news desk at the Hindustan Times, Mumbai, before joining Newschecker.
Ruby leads editorial, operations and initiatives at Newschecker. In her former avatar at New Delhi Television (NDTV), India’s leading national news network, she was a news anchor, supervising producer and senior output editor. Her over a decade-long career encompasses ground-breaking reportage from conflict zones and reporting on terror incidents, election campaigns, and gender issues. Ruby is an Emmy-nominated producer and has handled both local and international assignments, including the coverage of Arab Spring in 2011, the US Presidential elections in 2016, and ground reportage on the Kashmir issue since 2009.
