Your Children Will Thank You for Regulating AI

The example excerpt you used contained language that is actually very difficult to validate because it shifts with the rapidly evolving curve of AI research and innovation. Meeting it therefore requires a commitment to the use case, and model set, until it reaches stability. No one should be developing commercial AI who isn’t committed to having their models achieve stability in accordance with the current rate of the art for their target customer.

Working in a space that is rapidly evolving AND high risk requires a level of dedication that is counter to the ease of experimentation and exploration that enables people to create harmful AI solutions. Net net, if you aren’t committed to the AI industry, contributing to and achieving the evolution of its standard as the field matures, don’t do AI.

Furthermore the regulation itself admits that mistakes will happen, which makes it safe to assume only clearly negligent offenders of the policy will face real repercussions. The legislation is not designed to drive perfect AI, it’s designed to enforce the same standard of quality assurance that has been in place for other technologies but has consistently not been applied to AI and has therefore led to harm among populations at scale.

The people working on pushing these regulations aren’t fear mongers, they are responsible technologist, engineers, researchers, and activist responding to the crisis that has already been created by nefarious AI systems. Regulations aren’t created out of thin air because of imaginary fear — politicians don’t work that hard for nothing — these regulations were created in response to real incidents of harm at scale, not science fiction.

I hope the Arthur I am responding to you will have a change of heart and help AI practitioners evolve their development and quality standards to protect society rather than rage against the regulations designed to protect humans from harm at scale.