Advising clients at the early stages of starting a business can be tricky. Often, such clients view their corporate/commercial counsel as the legal equivalent of a Swiss Army knife. Discussions about corporate structure and agreements can swiftly lead to questions about HR, finance, tax, IP, regulation, competition, advertising, labour, real estate, etc. Sometimes, these demands are exacerbated by the complexity of the industries in which clients find themselves. That can certainly be the case when it comes to Artificial Intelligence (AI) and technology law, but understanding some of the basics and keeping a few key aspects in mind can help corporate/commercial advisors greatly.
What is AI and why am I hearing so much about it now?
When people talk about AI, they are usually referring to the general field devoted to creating machines (i.e. computers) having the capability of imitating intelligent decision-making in a given domain. In most cases, they are also referring to something called machine learning, which is a subset of AI and relates to the field of creating computers that can "learn". Such computers get better at doing something (e.g. playing chess) by seeing many examples of that thing (e.g. many chess games) and then building sets of decision-making rules that will allow them to react to given situations in future (e.g. if my opponent opens with move X, I counter with move Y).
The development of AI was predicated on having access to very large amounts of data (e.g. descriptions of every move and counter-move of over 700,000 grandmaster chess games) and extremely powerful computers (e.g. computers capable of evaluating 200 million chess positions per second). Until recently, these two constraints made AI a pursuit best suited to the Fortune 500. Our ability to generate ever-increasing amounts of data, as well as the advent of on-demand computing (cloud computing) and storage (cloud storage) has democratized the development and use of AI, and has led to it exploding into the business world.
What can AI do?
One of the reasons AI is so misunderstood is because its capabilities are simultaneously underestimated and overestimated. The ability of AI to solve particular problems is often underestimated, as people find it hard to conceptualise just how much data is being analysed by an AI, and just how fast that AI is analysing that data. It is also hard for us to comprehend how fast this area of technology is evolving. A good example is IBM's Deep Blue (the example alluded to above), a chess playing AI that beat Garry Kasparov in 1997 by studying over 700,000 chess games and evaluating 200,000 moves per second. Deep Blue was developed by IBM Research over a period of about 7 years. Thirty years later, Google's AplhaZero, a game-playing AI, taught itself chess in 24 hours (by playing itself, no less!), and was able to defeat all of the world's top chess-playing AIs, which had left human players in the dust many years earlier.
While this is undoubtedly impressive, AlphaZero is not "smart" enough to make a cup of tea, nor can it understand what chess is or why anyone would want to play it. As the depth of AI is often underestimated (i.e. how good a machine can be at doing one task when compared to a human being), the breadth of AI is often overestimated (i.e. how many different tasks a machine can do when compared to a human being).
What can AI not do?
AIs are categorized in a number of ways, one of these ways being the broad concepts of "strong AI" and "weak AI". While not particularly useful as a categorization, as you can fit all of the world's AIs into the latter of these two categories, it is useful conceptually. That is to say, AI as we know it is limited to specific problem-solving or reasoning tasks (weak AI). An AI having human cognitive abilities (strong AI) or capable of learning anything a human can, is still very much science fiction. Many researches today speculate that we are still decades away from strong AI, and others think that we might never get there.
Is this just hype?
Yes, and no. It is hard to refute that "smart" and "intelligent" are becoming over-used marketing prefixes, much like "e-" and "cyber" were twenty years ago, and "web" was before that. It is, however, undeniable that there are specific applications for which the use of AI is advancing the interests of humanity in a way that could never have been envisaged without it.
What do I need to be aware of as a legal advisor?
AI needs data, and lots of it. Clients must understand the privacy and confidentiality issues around the data they own and/or use. Privacy law in Canada is governed by a regulatory framework and the common law. The Personal Information Protection and Electronic Documents Act ("PIPEDA") establishes the basic rules governing how private-sector organizations must collect, use or disclose personal information in the course of their commercial activities in Canada.
AI is collaborative. Another key point to be considered is that AI projects are often multi-disciplinary and collaborative in nature. The development work required to build commercially-useful AI tools is complex and requires a wide range of technical and commercial knowledge along with immense and diverse data sets. As a result, AI development is more collaborative than many previous technologies.
One way of addressing the collaborative nature of the AI industry is through confidentiality agreements and trade secrets. While a confidentiality agreement is a useful measure for protecting trade secrets and other confidential information, it is by no means guaranteed to protect your client's technology.
Another way of protecting your client's technology is to file patent applications prior to your client disclosing their technology during the course of a collaboration. The patent system rewards inventors who file applications early with a comprehensive description of their invention in various configurations. It is not always necessary to have a working version of a technology to file a patent application, though filing a patent with an incomplete disclosure of your client's technology can actually be prejudicial to their interests.
Accordingly, taking some time early on with clients to consider privacy and confidentiality issues, as well as the right timing for seeking patent protection, can be hugely beneficial when advising AI start-ups.
The content of this article is intended to provide a general guide to the subject matter. Specialist advice should be sought about your specific circumstances.