About Blueprint for an AI Bill of Rights

The US White House Office of Science and Technology Policy recently released a “Blueprint for an AI Bill of Rights". Generally, the blueprint introduced five principles to the public: Safe and Effective Systems, Algorithmic Discrimination Protections, Data Privacy, Notice and Explanation, and Alternative Options.

The blueprint listed most things people worried about AI as it developed and popularized with incredible speed. Some areas have been focused on by the public a lot such as data privacy, and 'AI that replacing humans.’ This blueprint released by the White House, according to CNN, was been prepared for 1 year and can be called a greatly awaited one. However, this blueprint at the same time would disappoint many people, as it lacks specific orientation and executive power. Most of the words are established with 'should’ but lack actual implementation. In other words, it basically sculptured an idealized prospect while ignoring the contradiction that hid in details; it is an idealized utopia.

For example, when the blueprint introduced Data and Privacy, it list examples that individuals’ interests have been violated, such as the data collection under insurance and broker companies. Then in 'What should be expected of automated systems’ it introduced the idea of AI algorithms should be designed to avoid collecting unnecessary data as basic. Then in the section 'How these principles can move into practice’, the blueprint quoted the Privacy Act from 1974.

The question here is, we are facing privacy problems from AI as well as other technologies because the past Privacy Act cannot be satisfied and fully implement in today’s space-time environment. Therefore people are looking for governments to take action to provide an effective guide. Now, for example, there are machine-learning models that generate images such as DALL-E and StableDiffusion, which owns huge image databases. To track all original resources of the image would be neither inefficient nor unrealistic. On the other hand, eliminating such kinds of models that already exist in reason of the privacy argument would probably make the situation more complex. It is a fact that Privacy Act cannot be applied. And if the Privacy Act is useless in the area, so are the 'principles’.

The release date of the blueprint is also sensitive to people who care about politics as it was only one month before the mid-term election. As the document’s idea is very broad, and mostly favors the interest of civilians but seldom mentioned developers and businesses, it is hard to not suspect if it is another vote-gathering act.

In horizontal contrast EU‘s Artificial Intelligence Act, which defines the risk level of AI, the scope of application, companies’ responsibility, prohibited items, as well as punishment mechanisms, the White House’s principle is not even at the start point. While America might be advanced in technology, the fall behind of correspond rule might make people fall into more distressed situations.
While it is probably hard and ineffective to formulate a top-down principle to regulate AI-industry, it is worth learning the way that formulates different bills for each specific area of AI, such as 'remote biometrics, deepfakes, system applications for recruitment, chatbots, and some specific applications that should be banned.’ The subdivision would make the orientation clear, and it sounds more executable if the government creates detailed principles so that the state can handle the details.

In conclusion, the Blueprint brings a positive signal that AI-industry needs attention, but even with that America is far behind Europe. Hopefully, people can see the urgency from the comparison, as well as the gap that the ideal principle brings and the reality we are facing.

Reference Articles:

https://www.cnn.com/2022/10/04/tech/ai-bill-of-rights/index.html

https://www.weforum.org/agenda/2022/10/understanding-the-ai-bill-of-rights-protection/

https://www.whitehouse.gov/ostp/ai-bill-of-rights/what-is-the-blueprint-for-an-ai-bill-of-rights/

https://artificialintelligenceact.eu/the-act/

https://zhuanlan.zhihu.com/p/509818933

http://www.haxililiang.com/zhuanlan/jijieshuzi/2021-05-12/33597.html