AI Isn’t Smarter Than a Baby—Yet
SMRTR summary
Babies learn to understand the world far more efficiently than today's most powerful AI systems, and researchers want to know why. A new test called EgoBabyVLM, developed by Meta, Stanford, and other institutions, challenges AI models to learn from infant head-camera footage, and current models fail badly. Scientists believe building more baby-like AI could make future models cheaper, less energy-hungry, and better suited for robotics.
SMRTR provides this summary for quick context. The original article belongs to Wired.
Read the original article