Computers can do a lot, but a computer that understands “natural language”, puns, rhymes, riddles, ambiguities, even a double entendre? In short, the challenge is to make computers replicate all of these apparently simple human capabilities. But if you ask IBM’s top executives, they believe that the power of a Watson supercomputer (named after IBM’s first president, Thomas J. Watson), which played the popular quiz show Jeopardy in 2011, will be available in handheld devices within a decade.
IBM designed Watson as a question answering (QA) computing system, which could process natural language, retrieve information, reason and “open domain question answering”. It was capable of generating hypotheses, analysing and “gathering evidence” and performing complex analytics. The design of the machine was aimed at answering Jeopardy questions in three seconds. According to John Rennie, Watson’s capability enables it to process 500 gigabytes — about one million books — in one second. The machine’s ‘brain’ was filled with encyclopaedias, dictionaries, thesauri, articles from newswires, and literary works. Its ‘knowledge’ also included databases of all sorts, taxonomies, and ontologies.
What’s the need?ADVERTISEMENT
ADVERTISEMENT
Why would an ordinary user need a monster handheld for everyday tasks? For quite a lot, if you ask Bernard S. Meyerson, vice-president for Innovation, IBM Corporation. Speaking to The Hindu on the sidelines of the Association of Computing Machinery-International Collegiate Programming Contest (ACM-ICPC) in Warsaw, Mr. Meyerson says ‘big data’, and the way it is used, will change consumer behaviour significantly in the years ahead. “The game is not so much about the increasing volumes of data but about how to get the key trends that lie embedded in the mass of information,” Mr. Meyerson says, referring to the rise of analytics as a core area that lies at the intersection of big data and the power of computing.
“Our understanding of the evolution of computers and computing power has been powered by the application of Moore’s Law, that the challenge is to make computers smaller, faster and cheaper,” says Mr. Meyerson. “At this stage of the game, that is not only untrue, it is false. Most people do not admit it because it does not help their business,” he avers. Chips are now down to the atomic scale, he observes, pointing to the fact that IBM produces chips that are 5 to 7 atoms thick. “Since atoms do not like to be made smaller, there is a limit to how much you can reduce the chip by,” he notes.
He argues that technological progress in the field of computers in the future would not merely be more of the same old thing. This is not only true of memory and processors, but also for computer storage.
ADVERTISEMENT
Referring to discoveries made a few weeks ago, Mr. Meyerson notes that 12 atoms is the smallest bit that can be made on a disk drive. “If you try and make it 11 atoms, because of quantum mechanical fluctuations, the bit actually forgets whether it is a one or a zero,” he observes. That means the possibility of condensing the capacity of a disk drive is now limited to a factor of about 100 before it reaches the limits of possibility.
“Chips may get to 7 to 8 nanometres, but guess what, that is only three generations from today,” says Mr. Meyerson. What does this mean for technology companies? Mr. Meyerson says: “If all you do is disk drives, or all you do is chips, or all you do is memory, you are dead pretty soon.”
Cognitive computing
IBM, he says, aims to get a “holistic picture” of the world, because the next wave of computing would be in the area of “cognitive computing, which is an emergent technology”. Skills, especially programming skills, have never been more important, he says. “The technical trajectory of the past is pretty much done, and it will no longer be about getting one thing right,” he says. It is not about getting right chips, disk drives or software, but about all of these and more.
Douglas Heintzman, Director, IBM’s Lotus Strategy, believes the power of a Watson would be available in form factors that belong to cellphones, tablets and similar devices.
Mr. Heintzman claims that IBM has already deployed the power of such machines in several hospitals in the U.S., which enables the machine to engage in a “continuous dialogue” with doctors (including those at the Memorial Sloan-Kettering Cancer Center in New York) in order to develop practices that nurture “evidence-based medicine”. This means that it not only has an answer to a medical query, but is able to reason why a particular line of treatment is justified.
What this implies is that a machine like the Watson is constantly expanding the limits of what it knows, just like humans. “A doctor who has access to such a machine now has the opportunity to listen to the rationale and see whether the suggested course makes sense,” says Mr. Heintzman.
Mr. Meyerson argues that while humans are good at understanding causal relationships, they are much slower in understanding complex systems such as figuring out the traffic patterns in a mega city even when they have all the data.
For complex questions
For instance, how would the quantum of rain in a city like Rio de Janeiro affect traffic flows? Which areas are likely to get flooded and to what extent? These complex questions of everyday life require the marriage of diverse data sets, such as those on atmospheric conditions and traffic patterns on a real time basis, and in this case require the power of cognitive analytics, observes Mr. Meyerson.
Such ‘cognitive systems’ differ significantly from existing systems that are “driven by transaction speed and volumes”, argues Mr. Heintzman. The analytics of this kind is not merely about sifting data to get trends, but is fundamentally different in that the analytics “is cognitive in nature”. He predicts that a large part of the computer-related industry would be focussed on the area of “cognitive systems” area in the next 20 to 30 years.