Advancements in AI: From Capturing Human Intelligence to Simulating Human-like Agents
Originally Published 2 years ago — by Livescience.com
Neural networks have achieved a breakthrough in capturing a critical aspect of human intelligence, according to a study published in the journal Nature. Historically, neural networks have been criticized for their inability to combine known concepts in new ways, a capacity called "systematic compositionality." However, researchers have now developed a method called meta-learning for compositionality (MLC) that allows neural networks to practice applying different sets of rules to newly learned words, resulting in performance that matches or exceeds that of humans. While the model still has limitations in generalization, the study represents a significant step forward in training networks to be fully compositional.