Canadian research company MontrealAI convened leading computer scientists in December to debate moving artificial intelligence (AI) forward in the year ahead.
Cognitive scientist Gary Marcus cited key drawbacks of deep learning, including excessive data requirements, low capacity for inter-domain knowledge transfer, opacity, and a dearth of reasoning and knowledge representation.
Early last year, Marcus suggested hybridizing learning algorithms and rules-based software, while computer researcher Luis Lamb proposed a foundational strategy for neural-symbolic AI based on logical formalization and machine learning.
ACM A.M. Turing Award recipient Judea Pearl said AI systems require world knowledge and common sense to use the data they receive most efficiently.
View Full Article
Abstracts Copyright © 2021 SmithBucklin, Washington, DC, USA