Far from elementary, my dear Watson


Half a century ago, when scientists first convened at Dartmouth University to discuss the future of artificial intelligence, few of them thought that it would take this long to produce something that did a passable impression of a human being. For years, the closest we got to a machine that seemed smart was a computer that played chess. Now, IBM has pushed the envelope considerably, opening the door for artificial intelligence that will make a significant difference in vertical sectors such as healthcare.

Watson, IBM's latest supercomputer project designed to showcase its hardware and software smarts, won a competition against two human competitors while playing the game show Jeopardy in February. The machine, which interprets natural speech, processes questions against a huge database of information to produce a probability of a correct answer. The machine constantly refines its understanding of the various categories within Jeopardy using feedback on which answers were correct and which were not.

There may be a huge amount of processing work going on behind the scenes to give Watson the appearance of being human, but the results are impressive. The machine, which used 4 TB of disk space storing 200 million pages of content, processed approximately six million logic rules using 2800 processor cores.

Of course, Watson was not self-aware. We still have some way to go before we can create a machine that understands what it is. But doctors are unlikely to care about that. Watson is already planned for deployment in the healthcare sector. The idea is to use its expertise in handling huge tracts of unstructured information using natural language, and apply it to the vast archives of medical information at our disposal.

Trying to understand a conversation between a doctor and a patient and reduce helpful input may be difficult for a machine configured to provide very short answers to trivia questions. Medical information is highly specialized and nuanced, and can also be infuriatingly ambiguous thanks to misspellings and poor formatting. However some doctors hope that in a few years, the technology may have evolved to the point where Watson's offspring may be able to help doctors and nurses as a kind of digital co-worker.

In the meantime, some hope that it could be used to process calls at health centres. A clerical job seems a poor compensation prize, however, for a machine that took home $1 million for its creator from one of the most famous gameshows on earth.

Stewart Baines

I've been writing about technology for nearly 20 years, including editing industry magazines Connect and Communications International. In 2002 I co-founded Futurity Media with Anthony Plewes. My focus in Futurity Media is in emerging technologies, social media and future gazing. As a graduate of philosophy & science, I have studied futurology & foresight to the post-grad level.