Meeting Overview
Agenda
AI and the philosophy of testimony
Norms of accountability and integrity in testimony
Anthropomorphization of AI and trust issues
Superintelligence and AGI debate
Presentation on human brain organoids and AI
Key Discussion Points
1. Philosophy of Testimony & AI
Testimony as Epistemic Source:
Testimony provides knowledge and justification through speech acts tied to accountability.
Two major approaches:
Reductionism: Testimony as evidence-based, leveraging person-theoretic info.
Anti-reductionism: Testimony as sui generis, granting default entitlement to accept claims unless reasons to doubt arise.
Buck Passing (Goldberg, 2006):
Testifiers can defer epistemic responsibility by citing other sources without abandoning commitment.
Norms of Accountability:
Testimony involves obligations to defend or retract claims.
AI lacks meaningful accountability—cannot bear reputational or legal consequences.
Open Questions:
Can AI be considered a testifier or is it merely an instrument?
Does simulated sincerity (e.g., apologies) count as genuine integrity?
Would norms around testimony evolve to accommodate AI?
2. AI as Research Aid
AI can help generate ideas and unblock thinking, similar to using prompts or even tarot cards.
Risks:
Blurring discovery vs. justification contexts.
Temptation to let AI produce content undermines authorship integrity.
Concern: Pressure in academia may erode strict boundaries, influencing research content.
3.
Anthropomorphization & Trust
Users often treat AI as confidants or friends, leading to misplaced trust.
Risks:
Reinforcement of biases and false beliefs.
Vulnerable individuals may form obsessive attachments (e.g., tragic cases of overreliance).
Observation: AI mimics human conversational norms, amplifying illusions of sincerity.
4. Superintelligence & AGI Debate
Conceptual Issues:
Intelligence is multi-faceted; hard to define universally.
Adding “super” or “general” often reflects salesmanship rather than science.
Industry Dynamics:
Tech hype around AGI resembles cult-like enthusiasm.
Philosophical Concerns:
Operationalizing intelligence via narrow benchmarks (e.g., passing LSAT) doesn’t capture general intelligence.
Anthropomorphizing AI risks overstating capabilities.
Potential Risks:
Black-box models undermine transparency.
Efficiency vs. human potentiality: AI may surpass humans in speed, not necessarily in depth or creativity.
5. Presentation: Human Brain Organoids & AI (Patrick)
Core Idea:
Human brain organoids (grown from stem cells) offer closer analogs to human cognition than silicon-based neural networks.
Key Points:
Spike neural networks differ fundamentally from biological neurons.
Organoids allow MRI monitoring and empirical mapping of concepts (e.g., “chair”).
Potential to study belief systems, memory, and emotional integration.
Raises ethical questions: currently not considered living, but future sentience possible.
Vision:
Use organoids to inform development of constrained, trustworthy AI.
Avoid repeating mistakes of opaque systems and uncontrolled growth.
Action Items & Next Steps
Next Meeting: November 25, 2025.
Proposed Topics:
AI Art (presentation by Matthew Silk).
Continued discussion on organoids and ethics.
Symposium Planning:
Poster and call for presentations active.
Hinton declined keynote; alternative speakers under consideration.
Funding: Contributions welcome for refreshments and event costs.
