Why is an AI Constitution necessary?
As a practicing lawyer for over 40 years, Larry Bridgesmith maintains a deep reverence for the US Constitution. Despite our failure as a nation to achieve its goals and aspirations, it remains a model for curbing abuse and protecting human rights. But because we have ignored its demands for equal protection and due process of law, its fabric has frayed and its luster has diminished. Yet it remains intact to remind us of the limits of power opposing the power of "We the people". In addition, it takes courage and sometimes outrage to bring its principles back into focus.
In the US and around the world, questions of human rights in the face of the government's power to police its citizens are drawing millions of people into the streets to protest abuse. Similarly, many people question the ability of Artificial Intelligence to control human behavior and decision making. As a result, it is proper to review the needs and values of humans in relation to Artificial Intelligence. And, vice versa.
The World Academy of Engineering, Science and Technology is hosting its International Conference on Artificial Intelligence and Law this week. It was planned for Toronto, Canada. Covid-19 intervened. Accordingly, It it is now a remote conference.
Dr. Adel Elmessiry and Larry Bridgesmith published a research paper for the conference. It presents the often debated dilemma of how intelligent should we allow AI to become? Are we turning human rights and well being over to AI?
Is it time to "rage against the machines"?
Call for an AI Constitution
Dr. Adel and Larry have collaborated and called for the creation of a constitution of rights and remedies as AI continues to grow in power and influence. Clearly, roles and responsibilities must be established for AI to achieve its potential. For humans to receive optimal benefits from the creation of AI applications, there must be definitions of how both can relate to each other. Humans behaving without limits is anarchy. AI behaving without limits is autocracy. And neither serves the other very well.
In their paper, Larry and Dr. Adel set out the legal and technological rationale for an AI Constitution.
Interpersonal relationships thrive by balancing the concepts of identity, ownership, liability and applicable rules. Like humans, technologists define the work of AI in terms of both behavior and thought. And just like humans, AI functions in these two realms at less than an ideal level as well as at a minimally satisfactory level. Both humans and AI relate to others at the acceptable nexus of ideal and satisfactory. Bridgesmith and Elmessiry proposed a test for developing an AI constitution.
If AI functions within the sphere of acceptable thought and behavior, it should be qualified to benefit from the protections of an AI Constitution. Otherwise, it should be denied the status of person hood and its protections.
How to Develop an AI Constitution
Just like the US Constitution defines the ideal relationship between citizens and government, an AI Constitution must provide guidelines for human and AI relationships. As a result, Dr Adel and Larry propose a conceptual model to work from.
The US Constitution does not prescribe human to human interactions. Instead, it provides an enforceable framework for regulating those interactions. Likewise, an AI Constitution would not control human to human interactions. But it would define the limits of human to AI, AI to human and AI to AI relationships.
Numerous examples of human and non-human interactions provide ample grounds to extrapolate to human and AI relationships. For example, criminal and civil liability is imposed on corporations for violating laws and regulations. Likewise, immunity limits exist for military and police behavior exceeding lawful human encounters. Governments (federal and state) and their officers are not "above the law."
"Autonomous AI" now approaches reality. An AI Constitution ensures that beneficial outcomes can be expected as it develops. As a result, AI and the humans who are affected by it will be better served.
However, this is only a start. Join us as we explore the legitimate limits of AI and human interaction. We must establish a framework we can use to enforce those limits.