Latest
Celebrities & Pop CultureBritt McHenry’s Drink Toss at Wizards Game Cost Her a Job3 hours agoCelebrities & Pop CultureJenna Ortega TV Break: What Comes After ‘Wednesday’ Ends3 hours agoCelebrities & Pop CultureTom Skerritt Dies at 93: The Captain Who Outranked Ripley3 hours agoMoviesHalloweentown 5 Is Real. Now, Which Marnie Gets the Broom?3 hours agoMusicReservoir Sues Suno and Anthropic Over a $10 Song Machine4 hours agoGaming NewsGTA 6 Burger Shot Is Real, but the Burger Is Not New4 hours agoTravelDelta Medallion Status 2027: The Freeze United Didn’t Make4 hours agoRumors & LeaksMass Effect Universe Trailer Leak Shows the Game EA Killed4 hours agoAstrologyHow to Find Your North Node Sign and What It Means4 hours agoHome & Garden6 Essential Thanksgiving Decorating Mistakes That Can Make a Home Feel Cluttered7 hours ago
Gear & Tech

New Microsoft AI Study Suggests People Can Only Tell the Difference Between Genuine and AI Images 62% of the Time

Gemini AI, Microsoft/artificial intelligence/ OpenAI AWS
Image of Microsoft AI, Courtesy Mollie Dominy

Whether you love it or hate it, AI has taken the world by storm. Content generated by artificial intelligence is becoming more authentic-looking every day, to the point where it can be difficult for human eyes to tell the difference. In fact, a new study by Microsoft suggests people can’t tell the difference between authentic images and AI-generated images almost half of the time.

The Microsoft AI Study

According to PC Gamer, Microsoft’s AI for Good research lab conducted a study of whether or not people could tell the difference between genuine images and AI-generated content. There were approximately 12,500 participants, who were tasked with labelling around twenty-two images each. The images were selected from a “collection of 350 copyright free ‘real’ images” and “700 diffusion-based images using DallE-3, Stable diffusion-3, Stable diffusion XL, Stable diffusion XL inpaintings, Amazon Titan v1 and Midjourney v6.”

The Results

The Microsoft study found that the participants correctly labelled the pictures only 62% of the time. This means that the generators had a 38% success rate of fooling the participants. We’re getting past the days when all you had to do to identify an artificially-generated image was check and see if the hands had the correct number of fingers. Artificial intelligence is getting better at copying genuine art and images by the day. Can human intelligence keep up with it?

Try It Yourself

You can test your skills at identifying generated content here. I tried it, and I labeled just 33% of the images correctly, which, according to the site, puts me ahead of 3% of all the players. The sad part was that for most of the images, I seriously couldn’t tell the difference, and I resorted to blind guessing. As the controversy surrounding artificial intelligence continues to grow, we should possibly remind ourselves that not everything we see is real. Sometimes, a machine can fool us. The question is whether or not we’re okay with being fooled.

Total Apex Media Don't stop here Every Score. Every Story. One Home. Breaking sports, entertainment and news from across the Total Apex network, updated all day. Get the Full Story →