I picked up this book by Microsoft at the AI Summit 2018 (conference swag!). If it seems like flying to England is an expensive price for a book, get a copy for free at the links at the bottom of this page.
I’m not going to review the whole book but only focus on a couple of parts of interest to me. One is the foreword, by Brad Smith and Harry Shum. The second is Chapter 2, entitled Principles, Policies and Laws for the Responsible use of AI.
The other two Chapters are entitled The Future of Artificial Intelligence and AI and the Future of Jobs and Work. They are good reads but they aren’t really the focus of this blog.
Some of the statements made here are very much aligned with my view of the current legal response to AI. There is a constant focus on ethical principles and best practices, at least in the current literature, which is often to the exclusion of consideration of legal doctrine. The forward suggests this may be a deliberate stance. I don’t know if that is the case or if is simply that often it is not lawyers doing the writing and therefore legal doctrine doesn’t come up. In any case, the forward suggests that we need to focus on legal aspects across areas where AI interacts with society (globally) now:
“It will be impossible to address these issues effectively without a new generation of laws. So, while we can’t afford to stifle AI technology by aopting laws before we understand the issues that lie ahead of us, neither can we make the mistake of doing nothing now and waiting for two decades before getting started.”
I like that this consideration does not suggest that many current legal doctrines will fail in the face of AI (they won’t) but that we will need new laws (as an example it references the need to adapt employment laws, rather than to suggest that current employment law doctrine will fail entirely (it won’t). And also the foreword points out that AI law doesn’t exist now as a legal field. Both those points are behind the impetus for this blog.
The foreword also suggests that governance of AI should be consensus based. I’m not sure I agree with that other than as an overall principle that as a general ideal, everything should be consensus based. We certainly like to hear from everyone when considering new law, but consensus based governance on material laws is not likely to be acheivable. Second, there’s a bit of a schism in the discussion in that it calls for governments (presumably meaning national governments) to enact laws and regulations while at the same time recognizing the global reach of AI. To the extent that one accepts the premise that it is difficult to control AI across borders, legal considerations cannot be limited to national governments.
This Chapter sets out six principles for AI. Readers of this blog will know I have been collecting my own. Microsoft proposes:
• Fairness — AI systems should treat all people fairly
• Reliability — AI systems should perform reliably and safely
• Privacy & Security — AI systems should be secure and respect privacy
• Inclusiveness — AI systems should empower everyone and engage people
• Transparency — AI systems should be understandable
Lots of discussion follows each principle. Thoughtful discussion.
OK, so why don’t I just adopt these primciples into my Collected Principles of AI Regulation? I think as I considered them the answer is that really these are (and the chapter says so) principles for guiding the devlopment of AI rather than its regulation. For example fairness (of outcome), transparency and inclusiveness are not necessarily principles that we (obviously) codify into commercial law although we may hold them to be important.
There is another section to the chapter entitled Developing Policy and Law for Artificial Intelligence. It is a lot less granular in terms of setting out specific principles. It has some good thoughts in it, however, such as:
“We believe the most effective regulation can be acheived by providing all stakeholders with sufficient time to identify and articulate key principles guiding the development of responsible and trustworthy AI, and to implement these principles by adopting and refining best practices. Before devising new regulations or laws, there needs to be some clarity about the fundamental issues and principles that must be addressed.”
I completely agree. I also like that unlike the six specific principles that are enunciated to guide the development of AI, this section is less adherant to setting out specific guidelines. I think that makes a lot of sense in that Microsoft is at the heart of the development of AI so that it is appropriate for them to think about what should guide their products. Laws and regulation, on the other hand, are not a Microsoft product so a discussion is much more appropriate from them rather than a list of principles.
The section includes some subsections on Data, Responsible and Effective use of AI and Liability. I noted a thought in there about competition law and data, and how the two interact as data is concentrated. It is proposed that public data should be non-exclusive to AI. This is an interesting take away for me that in the end will deserve a lot more thought than a couple of sentences that it gets in the chapter. The Liability subsection calls for the use of existing negligence principles to handle liability claims regarding software. This is another take away for me to consider. I do agree that negligence and tort law work just fine, as I have discussed elsewhere. I did though, after reading in the book about the sophistication that AI can bring to processes through complex software, wonder a bit about how well our judges and juries are going to be able to understand where to place liability. The Transparency principle in the first part of the chapter notes that most people are not going to be able to handle a series of algorithms in order to understand how specific AI operates–how will judges and juries manage that?
Both the foreword and Chapter 2 are simply set out, easy to read and thoughtful. Anyone with time for a couple of cups of coffee and an interest in this area should invest in reading them. As far as considering laws and regulation of AI, I’d call it a nice introduction. Brew yourself a third cup as well and read the rest of the book.