← All stories
AI & Tech

Architecture search reveals transformers need less structure at trillion-parameter scale

Cognitive Revolution · Intelligence on the Edge: Liquid AI's Ramin Hasani on the Search for Device-Native Foundation Models · July 4, 2026
Architecture search reveals transformers need less structure at trillion-parameter scale
Cognitive Revolution
Cognitive Revolution
Intelligence on the Edge: Liquid AI's Ramin Hasani on the Search for Device-Native Foundation Models
"Scaling laws define architecture. When we are talking about Transformers being this revolutionary thing, we are talking about maximum scale. The reason why transformer architecture and attention mechanism is such a brilliant architecture is the fact that it is unstructured. There is no structure. The larger neural networks that you make into infinite size, the larger neural networks you make, the more you want them to become less and less structured. As soon as you start adding a little bit of bias in that architecture at scale, things become completely messed up."
Liquid AI's automated architecture search across models from 10 million to 72 billion parameters revealed a fundamental scaling principle: larger models require less architectural bias and structure, while smaller specialized models benefit from complex gating and nonlinearity. This insight challenges assumptions about optimal architectures and suggests transformers' dominance at frontier scale is mathematically justified.
From this episode
Cognitive Revolution
Cognitive Revolution

Intelligence on the Edge: Liquid AI's Ramin Hasani on the Search for Device-Native Foundation Models

August 3, 2026 · 1h 47m · 7 Egleze moments
Read episode summary and key points →

More moments from this episode

AI & TechLiquid AI warns hardware makers must build intelligence layer or lose to NVIDIAAI & TechAnnual smartphone market equals data center buildout creating trillion-dollar edge compute opportunityAI & TechLiquid AI ranks fifth in US for foundation model downloads with only 1000 GPUsAI & TechLiquid AI CEO reveals 12-neuron network can autonomously park cars
More stories More from Cognitive Revolution