The TWIML AI Podcast (formerly This Week in Machine Learning & Artificial Intelligence) What Does it Mean for a Machine to "Understand"? with Thomas Dietterich - #315
Nov 7, 2019
In this discussion, Thomas Dietterich, Distinguished Professor Emeritus at Oregon State University and expert in AI, dives into the nuances of machine understanding. He critiques current systems, advocating for a richer definition of 'understanding' in AI. The conversation touches on the hype surrounding AI advancements and the path toward artificial general intelligence (AGI), proposing that AGI should be viewed as a collection of specialized components. Dietterich also explores the philosophical implications of understanding through the Chinese Room argument, challenging perceptions of true comprehension.
AI Snips
Chapters
Transcript
Episode notes
Understanding is Not Binary
- Critics argue that AI lacks real understanding, but human understanding is also imperfect.
- Understanding isn't binary; it's a spectrum of completeness.
Narrow vs. Real Understanding
- AI systems exhibit narrow but real understanding within specific tasks.
- Focusing on "real" understanding hinders progress by setting unrealistic expectations.
Siri's Limited Understanding
- When Siri successfully calls Dan, it demonstrates understanding within that context.
- However, Siri cannot grasp deeper meanings or emotional contexts, showing its limitations.