The fast pace of growth in artificial intelligence (AI) algorithms, like deep learning, has transformed many applications. One of the major beneficiaries is the field of robotics. Advanced robots can be seen as embodiments of AI, and hence candidates for autonomy, self-repair, and even production of further systems, with or without AI. This brings legal issues into the limelight regarding accountability of any wrong being done, limiting what kind of actions can be undertaken by such creations, and so on. Asimov started this discussion with his famous set of four laws. But these are woefully inadequate in practice.
The book begins its journey here. As to whether AI is advanced enough to address such concerns, the book suggests that this may happen sooner than expected. Spread over eight chapters and approximately 400 pages, the book explores the myriad dimensions of this issue.
Starting with a brief historic introduction to AI, the first chapter moves quickly to the need for laws and regulations for AI-based entities. Chapter 2 dwells on what makes AI technology unique in demanding such legal provisions. The key insight is that AI makes moral choices beyond what is fed into it by designers and developers. For most man-made systems, once human input is over, behavior is predictable; this is not true for AI-based systems. The book takes the view that we are fast approaching the age of maturity for AI.
Chapters 3 and 4 deal with responsibility and rights, respectively, as they would apply to any legal entity. The theme of discussion is how to approach these topics, including private law, criminal law, comparison to animals, and so on, and whether to treat AI-based entities as “an object, a subject, a thing, or a person.”
Chapter 5 takes this notion further: should AI systems be granted a personality? Personality has strong legal implications. Saudi Arabia granting citizenship to the Sophia robot is a case in point. All this leads to the notion of a regulator to manage and enforce the legal system for AI. Chapter 6 is on building such a regulator and reviews current work along these lines in different countries. On the other side of this debate are the creators of such AI systems. Though the possibility exists of AI systems exceeding the mandate given to them, the ethics and education to prevent such runaway situations is worth thinking about. Chapter 7 examines this aspect. Finally, chapter 8 comes back to creation, the entities themselves, and the laws/rules that may be applicable, including mandating a kill switch.
The book is quite readable, though the legalese does get in the way sometimes. It discusses the issues in detail and includes loads of references as footnotes, as perhaps one would expect from such a “legal” book. A recommended read for anyone working in advanced AI or robotics. A final thought: I hope the scenarios envisaged do not materialize with such severity!
