I think there is a straightforward model for how to make sure AI continues to be developed as safely as possible. We can look at civil aviation industry which was being developed a century ago and how congress dealt with issues of safety at that time.  The FAA dates back to the Air Commerce Act of 1926 which was passed to regulate civil aviation after a period of barnstorming and mail flight accidents.

The early years of aviation was in all essence a  technology race.  Civil aviation initially had no rules, and it was being developed worldwide simultaneously.  Accident rates were too high in the early years of civil aviation. Instead of accepting the inherent dangers, congress did something about it.

The FAA evolved to be a federal agency that manages domestic safety of aviation equipment, pilots, maintenance, and all operations. It does so in direct partnership with companies. Contractors have their own FAA certified inspectors on staff, trained to ensure safety of all aspects of civil aviation. The FAA cooperates with equivalent bodies worldwide, including Europe (EASA), Canada (TCCA), China (CAAC) and others.  All are coordinated by an umbrella organization the International Civil Aviation Organization (ICAO) which is a specialized agency of the United Nations.

Civil Aviation is not perfect, but for the most part civil aviation has proven to be incredibly safe, and yet there are still accidents, from which we learn and improve.

I believe AI development and deployment needs safety protocols and oversight, as well as, international cooperation. Companies, and the world will continue to benefit from the safe development and deployment of this technology, just as we all have benefited from safe development and deployment of Civil Aviation.  There will be accidents (as has already occurred) and we’ll have to learn from them and improve safety protocols. The people closest to the development know what needs to be done, but they need leadership and they need to be accountable and held accountable.

Yes, slow it down and make sure we understand what we are doing with this technology, but it is past time for Congress to act… and they have a great model to consider for how to approach the problem of AI Safety.  The FAA.