Facts About Fiction: How not to fall prey to an AI-generated image for real

On July 24, Democrat Jon Cooper, who raised $1 million for Obama’s campaigns and also served as national finance chairman for the Draft Biden 2016 super PAC, shared the AI-generated picture on his Twitter profile. The image went viral and also found its way to other social networking platforms, with people claiming it to be a real image of Biden and Obama.

Barack and Joe trolling Republicans — love it!! 🩷🩷 pic.twitter.com/qgn9L2Sbs8

— Jon Cooper (@joncoopertweets) July 24, 2023

On the same day, a few hours later, Cooper clarified that the image was not real.

Although I think virtually everyone realizes this image was photoshopped (duh!), I want to make clear that President Biden and President Obama didn’t REALLY take time out of their packed schedules to wear pink suits for a Barbie promo. But if you still have any doubt, here’s the…

— Jon Cooper (@joncoopertweets) July 24, 2023

But what if he or some fact-checkers had not issued a comment or published an article stating the truth? How would a common man scrolling the feed find out that the image was indeed AI-generated before someone issued a clarification?

Here are the tools that we recommend to cross-check the image:

1. ‘Optic AI or Not’

This tool detects if the images are made using Stable Diffusion, MidJourney, Dall-E, GAN or Generated faces. All these create realistic images using the texts issued as prompts.

2. Maybe’s AI-image-detector

This app is a proof-of-concept demonstration of using a ViT model to predict whether an artistic image was generated using AI.

One thing about the AI tools that generate images is that they do make mistakes, even if they are evolving rapidly. Currently, programmes like MidJourney, Dall-E and DeepAI have their glitches, especially with images that show people. For example, the image of the Pope wearing a white puffy jacket that went viral in the month of March exhibited only four fingers on the Pope’s hand.

getting ready for a big week pic.twitter.com/moG8rMzVtg

— Amjad Masad (@amasad) March 27, 2023

Other things to look for in these images are:

• Discrepancies when it comes to body proportions
• Ultra-smooth skin, which also includes smooth hair and flawless teeth
• Background of images: The objects in the background can appear deformed, or AI programmes can even clone people and objects and use them twice. Sometimes the background of the AI image is also blurred

Apart from this, a Google or a Yandex reverse image search is always advised to check for the source of the image, but if that doesn’t work as well, observation is the key. Also, it is always advised to check the social media accounts of the people in question. For example, in this case, check the verified social media handles of Biden and Obama. Had it really had been them, they would have shared the image on their profile or would have at least retweeted it.

Also Read

San Francisco moves to center of robotaxi universe after California agenc...

What is Project IDX, Google’s new AI-integrated coding environment?

ChatGPT Plus too pricey? 7 websites that let you access GPT-4 for free

Google’s Brain2Music AI interprets brain signals to reproduce the music y...

It is very easy to make an AI-generated image, but it is also easy to spot such images through careful observation and critical analysis.