Machine Learning Street Talk (MLST) #58 Dr. Ben Goertzel - Artificial General Intelligence
29 snips
Aug 11, 2021 
Ben Goertzel, a leading AI researcher and CEO of SingularityNET, dives into the ambitious quest for Artificial General Intelligence (AGI). He critiques current deep learning approaches, advocating for architectures inspired by human cognition rather than mere brain modeling. Discussing the potential of SingularityNET, Goertzel highlights the synergy of cognitive methods and knowledge representation. He also explores the importance of integrating neuroscience insights to enhance AI development, raising thought-provoking questions about creativity, consciousness, and the future of intelligence.
AI Snips
Chapters
Books
Transcript
Episode notes
Goertzel's Early AI Views
- Ben Goertzel disliked the hand-coded expert rules approach of Good Old-Fashioned AI (GOFAI).
- He preferred experiential, interactive learning and pattern recognition, even during the rise of neural nets.
Cognitive Synergy and Resource Constraints
- Cognitive synergy arises from limited compute resources, where AI processes must reinforce each other.
- Different knowledge types, like sensory and declarative, require specialized handling, necessitating synergy.
Graphs vs. Vectors in AI
- Graphs are better for abstract cognitive operations than linear algebra used in most neural nets.
- Graphs efficiently organize knowledge for nonlinear transformations, unlike linear operations on vectors.
