AI is the wave of the future, or so it’s said. Indeed, with computer processing power doubling every two years or so, luminaries such as Elon Musk predict the arrival of the first superhuman AI as early as the 2030s. In the meantime, however, we have IBM Watson, perhaps the most versatile AI product ever invented—and one that is expected to revolutionise industry after industry much sooner than 2030.
What is IBM Watson?
In its simplest form, IBM Watson is a computer program that can answer questions in natural language, by mining its answers from a vast database using its unprecedented generalised intelligence. You might say it works something like Google, except that it’s far, far more sophisticated—instead of providing a list of documents, it understands the information it retrieves and answers the question itself rather than referring you to web pages.
In addition to natural language processing, Watson also incorporates information retrieval, knowledge representation, automated reasoning and machine learning technologies. And unlike Google, it’s not free to use—in fact, it costs millions of dollars to set up a full version of Watson. Nevertheless, Watson’s cost is well within the range of affordability for many medium-sized companies.
Watson’s database and applications can be modified or supplemented according to the needs of the company that uses it, rendering it versatile enough to be useful across a wide variety of industries. In 2013, IBM announced plans to share Watson’s API with software developers, allowing them to modify it using embedded apps of their own creation. In other words, it’s not a super-app—it works more like a Microsoft operating system for AI.
Watson was originally developed to answer questions on the TV quiz show Jeopardy! Although its performance was weak at first, it eventually became so skilled that it beat two legendary human Jeopardy! champions to win a $1 million first prize. This result surprised many people, even industry insiders.
The significance of this feat was perhaps greater than the significance of the victory of another IBM AI product, Deep Blue, when it defeated world chess champion Garry Kasparov in a six-game match. After all, chess proceeds within the confines of a narrow set of rules. Winning at Jeopardy!, by contrast, requires broad knowledge on a great variety of subjects, in a data environment in which information is unstructured and problems are often not clearly defined.
After winning on Jeopardy!, Watson began assisting with utilisation management decisions in the healthcare industry, a field which it now dominates. It has since been adapted to solve a wide variety of problems, ranging from economic development in Africa to advertising strategy for multinational corporations. Its use is expanding rapidly, as companies feel compelled to use it before their competitors do.
Progress So Far: The State of the Art
In short, Watson is designed to help solve the modern problem of information overload by parsing through tremendous amounts of information at superhuman speed, thereby allowing humans to concentrate on creative activities that, for now at least, only humans can do. It is difficult, however, to understand Watson’s potentially revolutionary impact without going into at least a bit of detail into how it actually works:
- The natural language processing engine tries to understand a query using the Slot Grammar algorithm.
- Watson then searches its vast database, much like Google does. Watson can search both structured databases and unstructured material such as newswires.
- Watson generates possible answers, and then performs a secondary search that boosts the confidence of some answers at the expense of other, less plausible candidates.
- Watson calculates a confidence score for each answer.
- Watson selects the answer with the highest confidence score and reports it.
Answering questions by mining a built-in database is not all Watson can do, however. Here are some recent innovations:
Watson Anywhere: Watson Anywhere is designed to operate on any private, public, hybrid or multi-cloud system, even if data is split between platforms.
Watson OpenScale: Watson OpenScale is an open AI platform for managing various types of AI, regardless of where they were developed.
Watson Assistant: Watson Assistant is designed to build conversational interfaces into devices and apps.
Watson Assistant Discovery Extension: The Watson Assistance Discovery Extension can discover previously hidden insights in unstructured data.
Trouble in Paradise
Two main challenges face IBM going forward—competition from other companies and bugs in its own system.
Competition: AI has continued to develop rapidly since Watson was introduced. In particular, deep learning, a technique not originally utilised by Watson, has emerged as the leading AI development algorithm. Over the last five years, deep learning has triggered vast improvements in question answering. IBM has not been sitting idly by -- since 2015 it has been adding deep learning to Watson’s capabilities.
Bugs: In early 2019, IBM issued a security alert concerning five bugs that arose from Watson’s use of Java components. Potential problems range from information disclosure to vulnerability to remote takeover and denial of service attacks. Although IBM has not released complete information on all of these bugs (presumably for security reasons), it has issued a patch that’s included in the latest version of Java Runtime.
The law never quite catches up with technology. With the exponential development of technology and the linear development of law, the gap between what’s going on and what the law can accomodate is growing wider every day. As the legal system struggles to catch up, novel issues are starting to arise.
How IBM Watson Will Change the Working Environment
Watson is expected to revolutionise the working environment in many ways, not all of them positive. There’s no doubt that AI will replace many of the jobs now performed by humans. Angst is particularly severe in the legal industry, where it’s feared that Watson and other AI counterparts will soon be able to perform legal research and write contracts better than a flesh-and-blood lawyer can.
In most cases, however, we are less likely to see job displacement than we are to see job modification, as AI partners with human intelligence to enhance dramatically work performance and productivity. Indeed, business research company Gartner. Inc. predicts that one in five workers will have an “AI co-worker” by 2022. The resulting increase in efficiency might partially offset job losses by lowering the cost of services and thereby increasing demand.
Employment Law Implications
Indirect discrimination: Even in a best-case scenario, job losses as a result of Watson and other forms of AI are going to be significant, even if these losses are offset by job gains in other sectors. It’s inevitable that job losses will hit certain groups harder than others. This will implicate UK employment equality law, which forbids discrimination based on age, disability, gender reassignment, marriage and civil partnership, race, religion or belief, sex, and sexual orientation
Older workers, for example, might suffer a disproportionate impact if they lose jobs in traditional industries while lacking the technical knowledge necessary to compete in emerging industries where new jobs are likely to be available. Likewise, hiring algorithms, already in use in some sectors of the economy, might advantage one group of people over another in seemingly arbitrary ways, whether or not the algorithm itself reflects the unconscious biases of its creator.
Disability accommodation: With AI creations like Watson able to execute voice commands, people who were formerly classified as “disabled” might now be able to work as efficiently as their able-bodied co-workers. A quadraplegic, for example, could potentially perform AI-assisted research just as efficiently as anyone else.
Employers, who are required to make reasonable accommodations for disabled workers to avoid discriminating against them, are finding that the scope of “reasonable” accommodation is expanding as rapidly as the technology that drives them.
Freelancing and the “Gig” Economy: Freelancers, rather than traditional employees, are going to occupy more that 50% of the UK labour market by sometime in the 2020s, or so it is said, and much of this trend is being driven by location-independent AI systems such as Watson. Freelancers, of course, enjoy few of the benefits of employment—they aren’t even subject to minimum wage laws, for example.
It’s likely that the bright-line distinction between “employee” and “independent contractor” (freelancer) is going to have to be reworked in a manner that grants freelancers at least some of the benefits that employees now enjoy. Freelancers may even have to be unionised, something that has already happened to a limited extent in Denmark. If this doesn’t happen quickly, expect social unrest.
Ethical Implications: Asimov’s Three Laws of Robotics
The ethical issues raised by AI technology are both extensive and profound. And where ethical issues arise, legal issues cannot be far behind. AI is simply going to have to be governed properly, and time is running out to make the necessary adjustments. It’s a never-ending process, for course, but this is one area of AI law where the law cannot afford to fall too far behind the cutting edge of technological innovation.
In this context, science fiction author Isaac Asimov’s well-known “Three Laws of Robotics” seem just as relevant now as they were in the 1940s:
Law No. 1: A robot may not injure a human being or, through inaction, allow a human being to come to harm.
Law No. 2: A robot must obey orders given it by human beings except where such orders would conflict with the First Law.
Law No. 3: A robot must protect its own existence, as long as such protection does not conflict with the First or Second Law.
Of course, these laws are not likely to survive unmodified. Law #1, for example, would seem to prevent the use of “robots” (AI) for military purposes. IBM, for its own part, seems to be leading the way in actually supplementing these laws, however. In fact, IBM’s AI policies include a concept that could be characterised as a fourth law, paraphrased roughly as “Unless absolutely necessary, AI won’t be designed to replace humans, but instead will be designed to collaborate with them.”
It may take time for the specific nature of any further augmentation to become clear. After all, what’s obvious in hindsight is often invisible until shortly before it arrives. It’s unlikely, for example, that Apple executives anticipated the need for updated distracted driving laws caused by the widespread use of iPhones. To some degree, we’re just going to have to take it as it comes. Who knows? In the not-too-distant future, we may even need to consider an “AI Bill of Rights.”