Machine Learning Street Talk (MLST) #56 - Dr. Walid Saba, Gadi Singer, Prof. J. Mark Bishop (Panel discussion)
14 snips
Jul 8, 2021 In this engaging discussion, guests Walid Saba, Gadi Singer, and J. Mark Bishop explore the future of AI beyond deep learning. Saba critiques the limitations of current statistical methods in conversational agents, while Singer emphasizes the need for hybrid models that blend reasoning with data. Bishop dives into the philosophical boundaries of computational cognition, challenging the notion that AI can replicate human understanding. Together, they advocate for a cognitive approach to align AI closer to human values and reasoning.
AI Snips
Chapters
Transcript
Episode notes
Limitations of Deep Learning
- Deep learning excels at pattern recognition but struggles with meaning and reasoning.
- We need AI systems that understand language, not just predict words, and integrate knowledge.
The Humanity Gap
- Predictions about AI capabilities have often fallen short, revealing failures in understanding.
- Computation alone may not achieve understanding, insight, or consciousness, creating a "humanity gap."
AI vs. Machine Learning
- Equating AI with machine learning is misleading, as learning alone won't achieve AGI.
- Knowledge representation is crucial, and much of it can't be learned from data.