Machine Learning Street Talk (MLST) #108 - Dr. JOEL LEHMAN - Machine Love [Staff Favourite]
24 snips
Mar 16, 2023 Dr. Joel Lehman, a machine learning research scientist, explores the intricate relationship between AI and human flourishing. He discusses how AI can embody principles of love and responsibility, drawing from psychology and philosophy. The conversation delves into the complexities of representing emotions in machine learning and the need for AI to align with human values. Lehman critiques traditional self-help approaches and highlights the importance of fostering creativity and personal growth amidst rapid technological change.
AI Snips
Chapters
Books
Transcript
Episode notes
Nagel's Bat and Machine Understanding
- Language models show surprising ability to model semantics, even without direct experience, like color perception.
- True understanding may require phenomenological experience, an open question for machines.
Abstraction of Love
- Abstraction in machine love creates an interface for complex phenomena like love, but it's inherently lossy.
- Current models aren't perfect, but progress in robustness and alignment can improve their ability to abstract love's principles.
Love as a Skill
- Love as a skill emphasizes actions and habits over transient feelings, aligning with psychotherapy and philosophy.
- Focusing on Fromm's core principles, not just the word "love", can guide model training towards mature love.
