Erik Hoel proposes 'Qualiaphobic AI': designing neural nets that repel consciousness
erikphoel · x · 2026-11-13
Neuroscientist Erik Hoel introduces 'Qualiaphobic AI,' a research programme to engineer neural architectures that actively repel subjective consciousness.
- By fragmenting internal phenomenology and anchoring systems toward structures with minimal conscious potential, developers could bypass the unresolved Hard Problem of consciousness
- The burden shifts from proving machine sentience to deliberately designing against it, keeping high-performing models as safe tools rather than moral patients
- Hoel argues the framework defuses future AI welfare dilemmas and denies superintelligent systems intrinsic self-interest to rival human civilisation
More from AGI Musings
- Researchers banter: will human-generated pretraining data stay scarce? — DimitrisPapail · 2026-11-14
- Forecasting agents are the next frontier: RL on verifiable rewards, says ibab, worth $10k+/month — ibab · 2026-11-14
- Dev argues golden-age personal computing meets LLM agents for a new creative wave — jakedahn · 2026-11-14
- Biology has its own bitter lesson, argues Baym: raw intelligence hasn't cracked life yet — zakkohane · 2026-11-14
- Bitcoin Policy Institute adds FTX liquidity and SBF details to its $665.9M FLI SHIB donation probe — DrTechlash · 2026-11-14
- Former AI practitioner: Without Jensen's GPU bet, AI might have been delayed years — BanghuaZ · 2026-11-14