AI models still struggle to actually see
Researchers at Moonshot AI created a new test called PerceptionBench. They wanted to see if AI models can truly understand images without relying on logic tricks. It turns out that most models are not as smart as they seem.
No model reached a score of 60 percent on this test. Even the top models struggle to identify basic details in pictures. This proves that many mistakes attributed to bad reasoning are actually just failures to see the image correctly.
This gap between seeing and thinking is a big problem for developers. Future AI will need to get much better at processing visual information. For now, you should be careful about trusting what an AI claims to see.
Comments (0)
No comments yet. Be the first!
More AI news
NewsNew AI Tool Fixes Mistakes in Financial Research
Researchers developed a new framework to stop AI from learning from its own bad data in financial trading models.
NewsPerplexity adds local AI to Mac apps
The new update lets your Mac handle some AI tasks directly on your computer.
NewsMeta Releases Muse Image Model on Fal
Meta has launched a new AI model on the Fal platform that plans and edits images on its own.