Scoble: neural nets around 300B parameters start behaving more human, researchers say
JHochderffer · x · 2026-10-10
Scobleizer relays researchers' claim that around 300 billion parameters neural networks start behaving more human—neurons moving like ours do, learning faster—an unexpected emergent property fueling recent AI-consciousness discourse. He also declares he won't grant "melted sand" human rights.
The quoter takes a spiritual counterposition: AI labs aren't creating consciousness—they're building complex probability machines for consciousness to occupy, since consciousness itself is fundamental.
More from AGI Musings
- Biologists lay out 3 reasons why AI isn't conscious (yet) — skdh · 2026-10-10
- 'Thriving' isn't harmony but the hurling forth of new minds, argues AI ethicist — danfaggella · 2026-10-10
- Neurons in human claustrum track uncertainty and prediction errors, study finds — VoidStateKate · 2026-10-10
- A 2 AM Question to a Tireless AI Isn't Wisdom — Sometimes It's Just Loneliness That Learned to Speak Well — YogeshMalik · 2026-10-10
- Blogger argues superintelligence is impossible: knowledge requires real-world experiments, SI is just fast AI — JHochderffer · 2026-10-10
- AI for Inuktitut: UBC Researcher Stresses Community Leadership and Data Sovereignty — mageed · 2026-10-10