AI Safety: Highlights Senate & Judiciary Committee

Some important topics:

(a) AI is currently manually (by a person) assessed for performance. There is a danger that AI learns to misguide the person that is assessing it.
(b) AI safety is expensive to private companies, and slows their AI software release dates, and therefore reduces their revenue.
(c) AI safety technology must be developed by government research departments as AI safety is not currently an important saleable commodity.