Expert systems for medical applications have been proposed and indeed have shown some measure of success over the years. I remember the principles of these cleverly indexed databases being used as long ago as the 1970s. That's almost equivalent to the Cambrian period in the evolution of the computer.
IBM have announced now that the computer system known as Watson that beat the best players of Jeopardy last year on U.S televisions is being re-purposed to absorb, index and understand medical texts so that it can bring to bear its vast capacity for sorting out wheat from chaff to the domain of medical diagnosis. The team in charge of Watson, or should this version be named "House"? is currently feeding it as many medical text books, case studies and even individual patient follow-up reports as they can in an effort to have this AI indexing and recognition system winnow out the grains of goodness in medical cases.
Right now, the software that defines Watson runs on a rack of ninety POWER7 processors that each have an eight-core processor running four threads per core to give a total of 2880 parallel processing engines. Jeopardy was a very public publicity stunt and proof of concept that was masterfully managed by IBM to bring the concept of intelligent computers to the masses. After all, the awareness raising exercise that was Deep Blue didn't exactly bring the cause of AI into the home of the common man. Jeopardy however was a bread-and-circuses coup that made anyone who was a fan of this esoteric and long-running game show format very well aware of what a computer could do.
Today, a top of the line smartphone has a four core processor and 32 gigabytes of memory on board. LG have already released a quad-core phone, My money is on the new iPhone and iPad getting a quad-core processor and Moore's law still holds true. This means that in twelve years we can envisage a handheld device as powerful as the Watson system is today. I'd like to propose a new law also. The Kurzweil Compression law that states that time for advancement in technology becomes more compressed as technology itself advances. According to the combined Moore/Kurzweil effect, that twelve years may shrink to ten or fewer.
A human being is as dumb as his lack of access to knowledge. On a conversational level, you or I could chat with and even hold a lively and interesting conversation with a Cro-Magnon person from forty-thousand years ago. Had one of us been brought up in a world as devoid of solid knowledge as they had, we too would be considered primitive and would not know anything about simple math or simple aerodynamics or simple electricity or cooking in an oven or making of cloth and so many other things that make us "civilised" If you took a person from that time and taught them to read, gave them a smartphone and a broadband connection then they too would be capable of confidently discussing aspects of the modern world that they discovered through that medium. In ten<?> years. Any person on the planet will have instant access to any aspect of knowledge that they desire to absorb.
Just like the Cro-Magnon person, there are so many things beyond our mundane experience that even having the concept of an idea or a technology or a novel use for treacle is utterly unthinkable. A case in point is the discovery of Graphene. Andre Geim, the Nobel physics prizewinner had the interesting idea of using Scotch-tape to pull the stratified layers of carbon atoms in graphite apart until he arrived at a single layer of atoms. Blocks of Graphite and Scotch tape have been around for many many years. Why didn't we have Graphene decades ago? Access to the wealth of ideas that inspire people to do things differently based upon what is known to be possible or thought to be impossible is where the Human race excels. There isn't enough room in any one person's brain for all that knowledge but an expert system in your pocket will give everyone a huge boost.
No comments:
Post a Comment