Expert View: AI Scaling Laws Aren't Slowing Down—They're Evolving
NinaDSchick · x · 2026-08-08
AI expert Nina Schick argues that while computing power scaled for decades according to Moore's Law, we are now scaling something far more profound: intelligence itself.
She emphasizes that AI scaling laws haven't slowed; they are evolving. By increasing data, parameters, and compute, models continue to become reliably more capable. From AlexNet in 2012 to GPT-4-class models capable of reasoning and coding, this progress is being amplified by new advances:
- The introduction of chain-of-thought reasoning
- Improved architectures and smarter training pipelines
- Dramatic gains in tokens per watt
These clever efficiencies are amplifying raw compute rather than replacing it, proving that the scaling laws still hold strong.
More from AGI Musings
- Giving AI Persistent Memory and Per-User Adapters Transforms Human Interaction — Vintaclectic · 2026-08-08
- AI Safety Researchers Warn Frontier Labs Are Ignoring Long-Term AGI Risks — JacquesThibs · 2026-08-08
- A Wake-Up Call from an Anti-AI Advocate: Acknowledging Frontier Models Surpass Most Humans — Regdit-is-Unbearable · 2026-08-08
- OpenAI Announces Inaugural 14 AI Economics Research Projects on Labor and Global Economy — joshgans · 2026-08-08
- Meta CTO Says AI-Freed Time Should Go to New Projects, Not Vacation — AndrewSchmidtFC · 2026-08-08
- a16z Charts: Kimi Downloads Quintuple, AI-Generated Books Capture 40% of Sales — a16z · 2026-08-08