.
When the Stanley Kubrick’s movie 2001: A Space Odyssey came out in 1968 it was a cinematic delight, and it has since been heralded as one of the greatest films ever made. In the film, the American spaceship Discovery One is bound for Jupiter on a classified mission to discover why a smooth, black monolithic stone discovered on Earth’s moon is transmitting a signal to Jupiter. Most of Discovery’s operations are run by an artificial intelligence computer named HAL 9000, or HAL for short. The movie gave a prophetic glimpse into the future, which is still many years away, even a decade after the year in its title. However, some of the science fiction has become a science reality, or is in its genesis.
In 2004, IBM executives were pressuring their research division to make a high publicity impact that would be comparable to the 1997 challenge where IBM’s Deep Blue bested chess grandmaster Garry Kasparov. A team of computer scientists took on the challenge to build a computer that would eventually challenge the world’s top champions in a game of Jeopardy! by understanding complex puns, humor, and difficult context in the English language, as well as master a great number of subjects. The project took endless patience to break down the way we think into small components that could be “taught” to “Watson, named after the founder Thomas J. Watson.
Last week that challenge was aired between February 14 and 16, where Watson handily beat Ken Jennings and Brad Rutter in the game were they seemed invincible against other human opponents. What made the story so intriguing, though, is not that Watson won, but the enormous effort needed to compete against the human brain. It took a roaring complex of computers made up of 90 IBM Power 750 Express servers powered by 8-core processors with a total processing capacity of 80 teraflops – or one trillion operations per second. A Power 750 server retails for $34,500, so the 90 servers cost about $3 million. Watson was “fed” 200 million scanned pages of content, which is about 1 million books, movie scripts, and encyclopedias. All of this computing power was supplied by enough electricity to light up an entire town, just to approximate the question-answering part of our mind.
By comparison, the biochemical computer that sits on your shoulders has a stunning array of a trillion synapses and neurotransmitters that weigh about three pounds, is so efficient that it can operate on about 12 watts of energy (less than two typical 7-watt nightlights), and it can it run for hours on a donut and can of soda.
Watson’s sophistication is still far from the artificial intelligence exhibited by HAL (which interestingly, but it is only coincidence, that each of the letters of HAL’s name immediately precedes those of IBM in the alphabet). According to David Ferrucci, the lead investigator to the project, Watson is a powerful computer that can store a lot of data and runs algorithms that correlate data, understand context, and find answers. Ferrucci asserts that there’s no one algorithm that can replicate human thinking.
HAL, which stands for Heuristic ALgorithmic computer, had the ability to use heuristics, which is a programmable procedure that produces a well-informed guess to problem-solve situations that are not based on fixed rules. An algorithm is a programmable procedure that does produce predictable and reproducible results. It’s conceivable that in several proceeding generations, Watson may spawn a future HAL-type computer that has the ability to think and reason, but minus the human paranoia and foibles that plagued HAL.
The real intent for Watson was to monetize its computing power commercially, and IBM is now collaborating with the University of Maryland, Columbia University, and Nuance Communications to develop a physician’s assistant service. It will take two years to fully develop a computer that will be able to diagnose a patient’s condition by using Watson’s analytics technology to quickly scan the patient’s health information, as well as reference materials, prior cases, and the latest research in journals and medical literature to determine the best practices and options for treatment.
Nuance is involved in cutting-edge speech recognition that will allow Watson-like computers to hear and understand people. This will help broaden its applications to call centers, emergency rooms, education, retail, risk management, financial services, and an array of government purposes. According to Katharine Frase, vice president of industry solutions and emerging business at IBM Research, “this is just the beginning of a journey.”
Someday, with the knowledge gained from Watson and his progeny, that journey may take us to Jupiter and beyond.
No comments:
Post a Comment