This post was featured in the Reality Defender Newsletter. To receive news, updates, and more on deepfakes and Generative AI in your inbox, subscribe to the Reality Defender Newsletter today.
Deepfake of the Week
The Pope recently made a public appearance covered in Balenciaga. Except, of course, he didn't, and this Midjourney 5-created deepfake fooled many online users at first glance.
What does Reality Defender's platform say?
Our platform found this to be fake in less time it took to upload on our web app.
How can users tell if this image is fake?
- Look at the hands: Generative images and off-the-shelf deepfaked images are particularly horrible at generating hands. Notice the Pope's right hand and its distortion.
- Look for text: Generative images are notoriously bad at rendering text. Look for text in images on signs, badges, and elsewhere. If it's a bunch of nonsense, it's almost certainly fabricated.
- Look at background faces: Images like this one often place emphasis on rendering the object or person(s) in the foreground, while the background assets are often a computational afterthought. The dimly-lit person in the background is, at first glance, convincing enough to be just a fuzzy bystander.
This instance shows the need for tools like Reality Defender to generate an in-depth forensic analysis and determine the validity of assets that could trick even the more astute observers. Leaving detection and validation up to users is a losing game. Deepfake detection and subsequent moderation on a platform level before media like this is posted is guaranteed to stop misinformation and deception before it hits a single user.
Bard Comes to Town
Bard, Google's hastily-launched answer to ChatGPT, is now live for select users. Like ChatGPT, Bing, and other lesser-known text generation tools, millions are already using Bard to answer questions, help with homework assignments, and give nuanced answers to incredibly complications.
Like the aforementioned tools, however, those answers might not always be correct.
A Deepfaked Earthquake
Do you remember the 9.1 Great Cascadia Earthquake & Tsunami that hit the Pacific Coast of the U.S. and Canada? Well, it never happened, but one Midjourney user created scenes from this horrible fictional event. Vice recently dug into Midjourney users' creations of false historical events and their deepfaked records.
How Reality Defender Helped NATO STRATCOM Detect Russian Misinformation
Reality Defender's advanced deepfake detection technology helped NATO STRATCOM identify deepfakes and misinformation spread by pro-Kremlin actors on VKontakte, providing them with the information they needed to make informed decisions and take appropriate actions to stop the spread of disinformation and propaganda.
More News
- A recent op-ed highlights the need for public policy to catch up to deepfakes and GenAI. (The Toronto Star)
- With deepfakes, anyone can be Batman. (Movieweb)
- Last week's Trump arrest deepfakes brought a whole new issue to light. (The Atlantic)
- Audio deepfakes are being used to clone your voice and trick others. (Gizmodo)
- Adobe has joined the text-to-image game. (Adobe)
- A reminder that it takes little time or effort to create a potentially damaging deepfake. (NPR)
Thank you for reading the Reality Defender Newsletter. If you have any questions about Reality Defender, or if you would like to see anything in future issues, please reach out to us here.
-The Reality Defender Team
