US needs nimble approach to technological governance
The new law comes as the feasibility of AI oversight is in question for being impossible and too expensive.
A new California law recently passed establishes a framework for independent verification organizations to test frontier AI models before their release, aiming to mitigate potential risks. However, the law's feasibility is questioned as being both impossible and too costly. As AI oversight becomes increasingly challenging, such laws may quickly become obsolete.
Frontier lab models have become so complex that they necessitate the analysis of other AI models. The METR investigation into the OpenAI-Hugging Face hack cost around $400,000 in AI tokens, a bill covered by OpenAI, and that was just one incident. Instead of legislation that attempts to freeze in place for rapidly evolving targets, a more adaptable approach to technological governance across the board is required.
We need rules and best practices crafted by knowledgeable individuals who are not afraid to admit when they are wrong and make corrections. Otherwise, we risk accumulating ineffective laws or, even worse, laws that produce the opposite of the intended outcome - something Californians are well acquainted with.
Written by urgent.news from Semafor's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
