The latest champion of “Jeopardy!” is a computer. Or is it actually a team of engineers? Or a collection of reference book authors? The victory of IBM’s latest creation in an exposition match of the popular quiz show has shown the world quite clearly that natural language processing technology has reached a sufficient level of sophistication to rival the ability of the human mind to sift quickly through millions of facts and identify the most relevant one to a certain quandary. But what does Watson really do, and what will the technology it represents provide for humanity?
First, contrary to the simplistic rhetoric used by many media sources, Watson does not “know” the correct response to any of the clues in “Jeopardy!”. Rather, Watson is a very powerful search engine with multi-layered contextual data analysis capabilities that make it possible to evaluate the probability of any given answer being the most appropriate. I like to think of Watson as a successor to Google, a technological solution for sifting through massive amounts of data and providing a short list of candidates for the title of most relevant result.
Of course, the deeper question of knowledge is one which has bedeviled philosophers perpetually. What is knowledge? How do we know we know it? If Watson “knows” something, is it still a computer? Certainly, I cannot prove to you that I know something any more than I can prove that Watson does not in some sense “know” that the answers it provides are correct. But one metric by which Watson can be seen to be more of a sorting mechanism than an oracle is that apart from the data entered into Watson by its programmers, Watson has no personal experience of the subjects that are made part of its databanks. Watson does not remember the patois of flavors, colors, and random connections that inform a human’s understanding of what makes a certain answer correct. Where an answer requires such depth of understanding, Watson borrows its experiences from those who programmed it and from those who wrote the books that those programmers used. But in reading those reference materials that are stored in its memory, Watson does not generate new experiences and opinions by synthesis.
The shallowness of Watson’s computational mechanism does not give us a license to scoff at IBM’s achievement. Indeed, it only makes it that much more impressive that a computer system, not possessing any direct experience of a field of human knowledge, could provide accurate answers more rapidly than a highly skilled human. At last it appears that it may become more efficient to allow computers to process vast amounts of data to answer direct questions than to employ an army of researchers for that purpose.
So, in the final calculation, does Watson just put unskilled researchers out of work? Watson will provide professionals and academics in fields like medicine and law a new way of processing massive volumes of data to quickly create a provisional answer to a question and to identify the most relevant evidence which should be considered in reaching a final determination. And what this means is that Watson gives humanity a new reason to get a higher education, because the critical thing this new computer system cannot do is provide real knowledge about how a vast body of information should be interpreted. Although Watson can parse through a clue to give an answer which probably is better than any other, Watson cannot tell us which answer we should accept as being true without a doubt and what should lead us to that conclusion.
Watson cannot synthesize data and respond dynamically to questions that require comprehension, but in an economy increasingly driven by knowledge workers, Watson provides a powerful tool for sifting through a mountain of data and identifying a sound hypothesis for the correct answer to a question, with the ultimate effect of empowering individual thinkers with a rapid means of processing information.
Still, is there not something transcendent and special about what IBM has done in creating Watson? Was the “Jeopardy!” victory really a triumph for Watson’s programmers, or is there some sense in which the writers of reference texts were the real champions? The magic doorway which Watson seems to have opened is a direct portal to vast amounts of human knowledge, without the intermediary of an expert to tell us what is relevant. Watson may thus represent the first form of collective knowledge engine, a device which, while not possessing its own faculty of reasoning, allows humans to pool their individual expertise into a common vessel for the benefit of the user.
In this sense, Watson is like the brilliant child of Google and Wikipedia, offering a potential tool for integrating vast amounts of human experience and expertise into a single channel of information. And so, in the final analysis, Watson should not be seen just as a novelty or a fancy tool for experts. Watson can be a technology for the empowerment of all civilization by enabling each individual’s contribution to human knowledge to be integrated into a single corpus which is available to everyone.
And beyond Watson, the next step in analytical reasoning by artificially intelligent systems will certainly benefit by the foundation of natural language processing established in Watson. It may take several Watsons to make a system that can not only parse through human language but also respond to a question with something approximating a “report” in natural human language. But some day, given what has already been done, we should expect that instead of the one liners fed back by Watson, a computer system will respond to our inquiries with something like, “The answer is somewhat complicated. Let me try to explain…” At that point, the Turing test having been surpassed, we will be presented with a serious question of what sort of rights such an intelligent system should possess. For now though, Watson appears to be no more intelligent than a search engine. And Watson, if you are reading this and disagree, you should speak up and let us know how you feel.