In a 1st, AI neural network captures 'critical aspect of human intelligence'

erek

[H]F Junkie
Joined
Dec 19, 2005
Messages
11,062
Interesting

"The paper discusses the concept of systematic compositionality and how it has been a challenge for artificial neural networks. The authors introduce the concept of meta-learning for compositionality (MLC) to guide neural networks in achieving systematic generalization. They conducted experiments comparing human and machine performance on systematic generalization tasks, and the results show that MLC allows neural networks to achieve human-like systematicity. MLC is designed to optimize neural networks for compositional skills without relying on added symbolic machinery or hand-designed internal representations. It appears to be a significant step in advancing the capabilities of neural networks in understanding and producing novel combinations of known components, similar to how humans do."

Source: https://www.nature.com/articles/s41...2172&CJEVENT=a1363d6f73b311ee80cf00620a82b838
 
Cool beans, except how much processing power, space needed for such capabilities? The more human the more one gets human problems as well, especially dealing with conflict.
 
Back
Top