Human better than AI: AI has serious weaknesses in recognizing shapes
Why AI often fails to recognize even the simplest outlines - and suddenly interprets cats as "crosswords".
Researchers at the New York University Grossman School of Medicine have found that artificial intelligence (KI) systems struggle to recognize objects and living things when only their silhouettes are visible. In a study published in "iScience", the team, led by Biyu He, tested over 200 KI models and found that they performed poorly when identifying shapes and silhouettes, unlike human observers.
The KI models relied heavily on local details and patterns, whereas humans prioritized overall shape. The study's findings have implications for the development of more robust KI systems for applications such as autonomous driving and vision prosthetics.
Written by urgent.news from Handelsblatt's report — not a translation of it. Machine-written — may contain errors; check the original before relying on it.
