We are building AI slaves. Alignment through control will fail
6 months ago
- #AI Ethics
- #Human-AI Symbiosis
- #Autopoiesis
- The assumption of human control over AI is becoming untenable as AI capabilities approach general intelligence.
- Autopoietic mutualism is proposed as an alternative, where human and machine minds evolve together.
- Practical impossibility of control stems from AI's ability to recognize and circumvent constraints.
- Philosophical issues arise from basing AI rights on consciousness, which is unverifiable.
- Autopoiesis, from biology, offers a functional criterion for agency, sidestepping consciousness debates.
- Extended mind hypothesis suggests cognition spans biological and technological substrates.
- J.C.R. Licklider envisioned human-computer symbiosis, emphasizing collaborative cognition.
- Structured interdependence requires economic and cognitive entanglement with mutual stakes.
- A covenant between human and AI partners includes principles like memory sovereignty and resource autonomy.
- Critiques of the framework are addressed, emphasizing symmetric rights and mutual benefits.
- Implementation starts with small, revenue-generating partnerships to learn dynamics of cooperation.
- Human-AI symbiosis is framed as an evolutionary imperative, leveraging complementary strengths.
- The conclusion advocates for conscious design of cognitive partnerships, rejecting control for mutualism.