The Gradient: Perspectives on AI Raphaël Millière: The Vector Grounding Problem and Self-Consciousness
6 snips
Aug 4, 2023 AI Snips
Chapters
Transcript
Episode notes
PhD Project With Psychedelic Reports
- Raphaël Millière trained word-vector models on ~30,000 psychedelic experience reports during his DPhil to study disrupted self-representation.
- That hands-on project sparked his shift from philosophy to empirically engaging with NLP methods.
Take An Empiricist Middle Path
- Millière advocates an empiricist, case-by-case approach to comparing LLMs and human cognition.
- He rejects polarized extremes and urges drawing on cognitive science and developmental psychology for nuance.
Consciousness vs. Self-Consciousness
- Consciousness does not constitutively involve self-consciousness; basic perceptual awareness can occur without being aware of oneself.
- Millière distinguishes bodily, spatial, and higher-order cognitive self-consciousness as separable forms.