March 28, 2006 11:39 AM PST
Sam Palmisano's technology forecast
IBM Research recently provided its Global Technology Outlook to Palmisano, who makes few public speeches and gives almost no interviews. The GTO is essentially a seven- to eight-hour presentation in which scientists from IBM's storied research division discuss what they think will emerge as major technology trends over the next three to five, or even 10, years. Ideally, IBM then comes up with ways to capitalize on the trends.
"It's about 35 to 50 slides per topic," said Mahesh Viswanathan, a vice president of strategy at IBM Research.
This year's GTO encompassed five topics. They were:
Moore's Law will prevail for at least 10 years, IBM researchers predict. Chip designers will have to incorporate new structures and chemicals into their chips, but they won't have to swap the silicon base with more radical materials for a decade or more.
Two years ago, IBM was less optimistic about the future of silicon, Viswanathan said, but the ability of engineers to keep shrinking transistors continues to surprise him.
Governments, established corporations and start-ups have all shown increasing interest in sensors that can more easily track the movement of cargo, cars or even people. But what do you do with all of the data collected by the sensors? IBM will likely start to look into ways of combining sensor networks with data mining (developed originally by Rakesh Agrawal at IBM). Conceivably, amassing the data from several sensor networks could enable researchers to better understand traffic patterns or the early warning signs of disease outbreaks.
Now that multicore processors exist that contain hundreds of millions of chips, it is becoming more economical to produce chips, or cores within chips, that perform specific functions. A significant market for math processors and other specialty chips existed years ago, but many functions got absorbed into general-interest microprocessors. The pendulum is now swinging the other way, because of the complexity of workloads and the large transistor budgets available to today's designers.
Server accelerator chips like those from Azul Systems and Japan's Institute of Physical and Chemical Research are a type of application processor.
Everyone makes software
In the past, a few people inside corporations wrote code. "Now everyone is a programmer," Viswanathan said. IBM and others will have to come up with tools to make writing programs easier, but also tools to ensure that these ad hoc applications can play well in existing corporate environments.
Everyone needs services, and IBM has tied its future to it. One of its big aims is to more acutely study the way organizations behave. Some at IBM acknowledge that it sounds like squishy science--but computer science wasn't recognized as a discipline when it first emerged, either. Stanford University's engineering school initially didn't teach it.
2 commentsJoin the conversation! Add your comment