Watson as a Warning: Current Limits and Promise of AI
We can’t seem to decide whether AI will replace humanity as the dominant intellectual force or it will simply be a sideshow never reaching its promise. Is AI the technology of the future, or will it always be so?
The answer is likely somewhere in between.
Bill Gates once noted that transformative technologies are overestimated in the short run and underestimated in the long run. AI is certainly likely to be such a transformative technology.
Last week the New York Times ran a cautionary tale about how IBM tried and failed to use its heavily branded AI solution called Watson to transform the business world and solidify IBM’s future. The Times noted that, in the years since Watson’s famous Jeopardy win over Ken Jennings, IBM’s stock has fallen ten percent while stock of AI/Cloud competitors like Amazon and Google burst into the stratosphere.
Changing the world with one product is more difficult than senior management at IBM anticipated. The Times wrote “The company’s top management, current and former IBM insiders noted, was dominated until recently by executives with backgrounds in services and sales rather than technology product experts. Product people, they say, might have better understood that Watson had been custom-built for a quiz show, a powerful but limited technology.” A tool custom made for one job can’t always be easily converted for other jobs.
At the moment, our only AI tools fall into the category of “narrow AI” – tools designed to perform a function or limited groups of functions in a certain space. Siri may seem like she knows everything, but she doesn’t. She was designed to interpret speech in specific languages, to access databases and Apple’s platform features, and to respond accordingly in the (admittedly impressive) limited circumstances she is likely to encounter. AI can shock us with speed and apparent breadth of knowledge in chosen tasks that reach far beyond what humans are capable of, but only for a specific set of actions.
AI can shock us with speed and apparent breadth of knowledge in chosen tasks that reach far beyond what humans are capable of, but only for a specific set of actions.
This was part of IBM management’s problem. They knew that Watson was “smart” – she could beat two Jeopardy champions – and that such skills could simply be applied to resolve the big problems in medicine, finance and government. Maybe an AI will be able to do so someday, but it will resolve those problems when it is built to resolve those problems. IBM has created an excellent narrow AI solution in Watson, but IBM assumed it had a general AI – an AI that could resolve any problem. General AI only exists in science fiction, and is unlikely to exist in our world any time in the near future.
Where IBM is correct about Watson, and how Watson is likely to be a foundation of IBM’s business over the next 50 years, is if IBM can use Watson-like AI to resolve sets of specific business problems, then IBM will have built a business solution that can be resold unlimited times to solve similar types of problems. Replace a convoluted business process for one bank, then you can repeat the process for all other banks at a high margin. And, if the business process is general enough, then maybe your solution translates to many other kinds of companies and government agencies. This is the Microsoft model, except with AI rather than old-line software. Serious profits ensue.
But the problems approached must be safely within AI’s capabilities which are likely to involve grouping, including or excluding samples from the groups, and applying logic to huge datasets. Michael Jordan proved to himself and the world that while he may have been the greatest basketball player of his era, he couldn’t simply translate that athleticism to other sports. We have never seen a human that is the best athlete at all sports, the smartest scientist, the greatest public speaker, and the winningest Jeopardy champion. Why does IBM expect that it can create an AI equivalent? Why would it need to?
We have never seen a human that is the best athlete at all sports, the smartest scientist, the greatest public speaker, and the winningest Jeopardy champion. Why does IBM expect that it can create an AI equivalent? Why would it need to?
IBM’s new management has recognized this truth and embraced it, to profitable effect. The Times articles states, “Now, Watson is a collection of software tools that companies use to build A.I.-based applications — ones that mainly streamline and automate basic tasks in areas like accounting, payments, technology operations, marketing and customer service. It is workhorse artificial intelligence, and that is true of most A.I. in business today. . . IBM says it has 40,000 Watson customers across 20 industries worldwide, more than double the number four years ago.” Watson may still be the foundation of IBM’s future business, but it will do so by applying its limited strength to specific problems.
A recent McKinsey report proposes that about half of current work activities are able to be automated with AI and other technologies. These include data collection and processing, as well as physical tasks in structured environments. So it is likely that AI, which is already transforming the workplace, will be seen as a major force for business and government efficiencies in the near future. And the education and maturity that comes with learning the strengths of this technology will allow AI to develop into its next level.
Will we ever produce a general AI that can accept input in dozens of different fields and spill out answers to our most profound problems? For better and worse, we won’t have this technology any time soon. But it is unrealistic to expect it. Watson is a spectacular technology for meeting the challenges it was designed to overcome. We need to recognize its potential productive uses, and not get lost in dreams of a panacea. Because, as science fiction tells us, a general AI can generate its own set of problems.