Nowadays, many online pictures are generated by AI, and they are not just ordinary pictures, but also design materials, such as Recraft It can generate SVG vector images, so can we identify whether the image is generated by AI? Many people will definitely want to challenge, in addition to the “AI real person recognition test”, Google has also launched the Odd One Out online test to test your eyesight, and it is more difficult. Choose 1 out of 4 pictures generated by AI, and there is a 4-life mechanism, which means that it will be over as long as you choose the wrong one 4 times.
Odd One Out is an AI picture-finding game launched by Google. Can you tell the difference?
Odd One Out is a small game in Google Arts and Culture. It does not currently support Chinese, but the 4 pictures that appear are all of the same type, so it doesn’t matter if you can’t understand English. Just click on the picture to start the experiment:
Each time you play, you have 4 chances to make the wrong choice. If you exceed the limit, it will stop. If you keep making the right choice, it will continue to see how many points you can get:
Four pictures will appear every time, allowing you to choose which one was generated by AI, and there is a time limit. There is a time bar at the bottom. If you don’t find it after the time, it means you lose, so you can’t take your time to look at it. You have to identify it quickly:
After answering correctly, the source of each picture will be explained below. This is quite good. You can absorb knowledge and explain the famous paintings and the authors. Below the AI image, only the generated image will be marked:
After all 4 lives are used up, the Game Over interface will appear. Here you can browse the questions you answered correctly. The lower left corner will also tell you how many points you got, so you can share it with your friends:
I have to say that every question is super difficult. Who can identify which picture is a picture of a dog generated by AI? However, there are still some tricks. Usually the pictures generated by AI do not have complex backgrounds, that is, there is only one subject. However, of the 4 pictures that appear each time, more than half of them are often a single subject:
It seems that the highest I passed the test was 4 times, which was more difficult than before. AI real person recognition testIt’s still high. I feel that AI portrait pictures are easier to identify than pictures of objects and animals. People who can get high scores in this test really admire it.
According to the description, the game was developed by Caroline Buttet and Emmanuel Durgoni, two artists-in-residence at the Google Arts & Culture Lab.
The way the game works is that players have to identify an AI-generated image from four different artworks in a short period of time. Every time they guess wrong, the player loses a heart, and every time they guess right, the answer time becomes shorter and shorter. This game also indirectly proves that AI drawing platforms like Midjourney and DALL·E have made great progress in creating images of some works of art.
Source: KOCPC Chinese