The sweeping AI safety bill just passed the California state Assembly yesterday. **Andrew Ng** (the author of this article) is a pioneer in AI research and has been vocal against this bill since it was first drafted.
The problem with the bill is it aims to regulate the **general-purpose technology rather than the applications of said tech**. For instance if an LLM model released by Meta and an X user used that model to harass elderly people on the telephone all the blame will be put on Meta.The bill is specifically more harmful to **open models because they are the most likely to fall to bad downstream use cases**. The availability of open models is **critical to advancing AI safety research** which is the whole goal of SB 1047.
The bill will also put a lot of **compliance burden on small developers and researchers** which will make it hard for them to compete against big companies and stifle the former in innovating new cutting-edge AI technology.
Finally and something not mentioned in this article is the **possible proliferation of AI regulation laws between the 50 states** if this Bill is signed into law by CaGovernor Newsom. A complex regulatory market makes it very hard for the US to seamlessly lead the world in AI progress. That’s why while I’m anti government intervention I would rather have AI regulation at the federal level.
1 Comment
The sweeping AI safety bill just passed the California state Assembly yesterday. **Andrew Ng** (the author of this article) is a pioneer in AI research and has been vocal against this bill since it was first drafted.
The problem with the bill is it aims to regulate the **general-purpose technology rather than the applications of said tech**. For instance if an LLM model released by Meta and an X user used that model to harass elderly people on the telephone all the blame will be put on Meta.The bill is specifically more harmful to **open models because they are the most likely to fall to bad downstream use cases**. The availability of open models is **critical to advancing AI safety research** which is the whole goal of SB 1047.
The bill will also put a lot of **compliance burden on small developers and researchers** which will make it hard for them to compete against big companies and stifle the former in innovating new cutting-edge AI technology.
Finally and something not mentioned in this article is the **possible proliferation of AI regulation laws between the 50 states** if this Bill is signed into law by CaGovernor Newsom. A complex regulatory market makes it very hard for the US to seamlessly lead the world in AI progress. That’s why while I’m anti government intervention I would rather have AI regulation at the federal level.