Two and a half years ago, in early 2011, an IBM-developed supercomputer named Watson won a Jeopardy! tournament against two human opponents. It was simultaneously fascinating (computers are cool!) and terrifying ("I, for one, welcome our new robot overlords."). Until now, however, the high-level computations performed by Watson have been kept locked away, but on Thursday, IBM will announce that it is making Watson available to companies and individual developers through the Internet.
The initiative is IBM's attempt to get into cloud computing services to compete with the likes of Google and Amazon. Rather than having to pay for hardware, software, and upkeep on complex technology, Watson's system of machine learning would be accessed remotely and, more importantly, could be far more inexpensive as well. The technology is already being tested out for medical purposes.
The point of Watson and similar existing cloud services is to distribute processes and functions across a wide network of servers rather than a single one. The Times report announcing IBM's new effort gives the example of an Amazon project that was completed in 18 hours on its network as opposed to 264 years on a single server. Obviously, that ratio is fungible based on specific technical specs, but the sentiment is still appealing.
IBM says that the API for querying Watson is simple, allowing users to pose questions in natural language, much in the same way they would talk to a program like Apple's Siri or WolframAlpha.
The upsides to this type of computing system are numerous. For one thing, IBM gets to keep their proprietary technology close to the vest, and as a machine learning system, Watson should more efficient and accurate as its dataset gets larger and larger. Customers constantly feeding it data points should serve to strengthen the product's language-processing and question answering abilities.