Neuro-Symbolic A.I. Algorithms and Systems

Companies Manipulating A.I. Systems for Competitive Advantage
March 10, 2020
Multitask Learning
March 10, 2020

Neuro-Symbolic A.I. Algorithms and Systems

The development of A.I. has been on two conceptual tracks since the 1950s: symbolic and non-symbolic.

The development of A.I. has been on two conceptual tracks since the 1950s: symbolic (machines use a base of knowledge and rules that represent concepts) and non-symbolic (machines use raw data to create their own patterns and representations of concepts).

Classic A.I. is the former, because it more closely represents how we understand human thought—and the original intent was to teach machines to think like us. Researchers are working on new ways to combine both learning and logic using neural networks, which would understand data through symbols rather than always relying on human programmers to sort, tag and catalogue data for them.

Symbolic algorithms will aid the process, which should eventually lead to robust systems that don’t always require a human for training.