Watson, IBM's supercomputer made famous three years ago for beating the very best human opponents at a game of Jeopardy, now comes with an impressive new feature. When asked to discuss any topic, it can autonomously scan its knowledge database for relevant content, "understand" the data, and argue both for and against that topic.
Watson's DeepQA is arguably the world's best computer system at natural language processing by a wide margin, which is an extraordinarily complex field of artificial intelligence. Perhaps the major difficulty in understanding human language is the lack of "common sense" in today's computers. For all its number-crunching power, Watson cannot "understand" the questions it is asked, at least not in a traditional sense. The way in which Watson answers questions is closer to symbol manipulation than to the way you and I understand and process information, but the end results are often impressive.
UPGRADE TO NEW ATLAS PLUS
More than 1,500 New Atlas Plus subscribers directly support our journalism, and get access to our premium ad-free site and email newsletter. Join them for just US$19 a year.UPGRADE
Watson looks at the question it is being asked and groups words together, finding statistically related phrases. Thanks to a massively parallel architecture, it then simultaneously uses thousands of language analysis algorithms to sift through its database of 15 terabytes of human knowledge and find the correct answer. The more algorithms find the same answer independently, the more a certain answer is likely to be correct. This is how, back in 2011, it managed to win a game of Jeopardy against two human champions.
In a presentation at the the Milken Institute Global Conference, IBM senior vice president and director of research John Kelly III demonstrated how Watson can now list, without human assistance, what it believes are the most valid arguments for and against a topic of choice. In other words, it can now debate for or against any topic, in natural language.
In a canned demonstration on stage, Watson was asked to present arguments for and against the sale of violent videogames to minors. After scanning Wikipedia for relevant information, Watson answered:
"I would like to raise the following points in support of the topic. Exposure to violent videogames results in increased physiological arousal, aggression-related thoughts and feelings as well as decreased social behavior. In addition, these violent games or lyrics actually cause adolescents to commit acts of real-life aggression. Finally, violent videogames can increase children's aggression.
On the other hand, I would like to note the following claims that oppose the topic. Violence in videogames is not causally linked with aggressive tendencies. In addition, most children who play videogames do not have problems. Finally, videogame play is part of an adolescent boy's normal social setting."
The answer wasn't just a mindless collage of sentences copied from various Wikipedia articles. Rather, in just a few seconds, Watson searched its sources for relevant information, scanned for arguments in favor and against the topic, selected what it believed were the strongest arguments, and then constructed sentences in natural language to illustrate the points it had selected.
For a computer that doesn't actually "understand" the questions it is being asked, this is a truly impressive achievement.
IBM believes that the technology behind Watson will prove very valuable in dealing with the exploding amount of information that we're currently facing. A fully automated system that can process huge amounts of data, extract information and find answers with a high degree of confidence could prove useful in a number of fields of human endeavor.
For instance, the system could have important applications in the medical arena. Oncologists could take a DNA profile from cancerous tissue, compare it to healthy tissue of the same organ, extract the mutations, and then use Watson to search the entire medical literature to find which specific combination of drugs will be best at targeting that specific mutation affecting that specific organ.
Back in February, IBM also announced it intends to use Watson to help countries in Africa find the answers to their development problems, with a focus on healthcare and education.
Watson is built on commercially available 750 Power servers, because IBM aims to market it to corporations in the future. The hardware to operate Watson at its minimum system requirements currently costs a relatively modest one million US dollars, but the price is expected to drop in the coming years.
The video below shows the new debating feature in action. The presentation starts at the 35 minute mark, the canned demonstration 46 minutes in.