Why compute-based AI regulations are becoming obsolete: Three key challenges
By
Séb Krier
Summary
This article examines the growing inadequacy of using pre-training compute as a proxy for AI model capabilities in regulatory frameworks like the EU AI Act. It argues that while pre-training compute was once a reliable indicator (accounting for 90-99% of total training compute), the landscape has shifted dramatically with modern AI development. The piece explores how compute-based AI policies must evolve to account for new training paradigms, post-training enhancements, and the changing relationship between compute and capabilities.
Source
Twitter / XWhy compute-based AI regulations are becoming obsolete: Three key challengesepoch.aiKey quotes
· 5 pulledWhen the EU AI Act was drafted, pre-training compute was a reasonable proxy for model capabilities.
At the time, pre-training accounted for 90-99% of total training compute, and the relationship was relatively reliable.
This simple proxy has been steadily breaking down.
While pre-training compute remains a primary driver of capabilities, modern AI development...
'Training compute' is constantly evolving, and compute-based AI policies must adapt to remain relevant
You might also wanna read
"Right-to-Compute" Laws Spread as States Take Opposite Approaches to AI Regulation
The article discusses the emerging legislative trend of "right-to-compute" laws that protect AI and computing from regulation, starting with
vktr.com·4mo ago
Examining the Limitations of Transformer Models and the Gap to Human-Level AI
The article presents a skeptical perspective on claims about imminent Artificial General Intelligence (AGI), arguing that current transforme
dlants.me·4mo ago
The EU AI Act's Operationalisation Gap: Strong Rules, Weak Infrastructure Control
The article argues that the EU's AI Act suffers from an "Operationalisation Gap" — while it establishes strong regulatory authority (Layer 1
The next AI infrastructure race shifts from chips to agent-ready systems
Wall Street has focused on compute power (chips, data centers, cloud infrastructure) as the primary lens for valuing AI investments over the
bit.ly·16d ago
Data Scarcity as the Emerging Bottleneck in AI Scaling and Intelligence Development
The article discusses the asymmetry between compute and data growth in AI development, arguing that while compute capacity grows rapidly, da
Neoclouds, sovereign AI, and Postgres: A new enterprise architecture for regulated AI inference
The article examines how AI inference is reshaping enterprise computing, particularly for regulated industries. As inference becomes dominan
bit.ly·8d ago
Comments
Sign in to join the conversation.
No comments yet. Be the first.