Today’s largest LLMs consume hundreds of thousands of watts of power and still fail to solve more than ~20% of ARC-AGI tasks. In contrast, the human brain runs on just 20 watts while solving over 90% of them. This stark efficiency and capability gap highlights that AI still lacks the human-like ability to approach problems with flexible reasoning. Any advancement, however small, in closing this gap would have profound implications across domains where few-shot learning, generalization, and discovery are essential.
Contact: office@scenda.ai
© 2025 Scenda.AI