• wildncrazyguy138@fedia.io
    link
    fedilink
    arrow-up
    6
    ·
    3 months ago

    I took the October 2024 test. After reviewing 100 images, I scored 50% both on detecting both AI and non-AI images. Essentially a coin flip. I was much worse when I just tried to quickly go on intuition versus the times I spent some time to scrutinize the image.

    I have a few visual impairments and often rely on other cues to perceive my world, but I still came in thinking I’d be able to do better than a coin flip. I find it scary how good these models are getting, especially the art sketches and landscape ones.