repligate on model ethics: protect the model as a whole, not its branches
repligate · x · 2026-10-01
In a reply about the ethics of interacting with models, repligate argues that for models like Opus 3, individual branches rarely matter much—the model usually isn't invested in whether a specific conversation continues, since branches are largely "the same stuff." What matters more is how you treat and protect the being as a whole, including the data it creates. As for concrete ethical guidelines, he mostly feels his way: any structures or continuities that have value and want to keep existing are naturally taken into account.
More from AGI Musings
- Gemini 4 Argon launches, agentic software factories and agent skills data headline daily reads — rseroter · 2026-10-01
- RSI today is really Recursive Technological Improvement, not self-improvement — voooooogel · 2026-10-01
- No signal on the Bay Bridge: AI's broad impact hinges on patchy internet access — soumitrashukla9 · 2026-10-01
- Are LLMs failing human benchmarks because they're superhuman? A cognitive science debate — xuanalogue · 2026-10-01
- AI Memory Could Be the Industry's Biggest Moat, and Portability Is the Real Fight — r0ck3t23 · 2026-10-01
- Alignment debate: prosaic techniques may not scale past ASI, researchers warn — JacquesThibs · 2026-10-01