January 13, 2011
This is just a personal reflection, but if there are two things on television that never get boring, even if I have seen the same episode multiple times, it's Star Trek: Next Generation and Jeopardy.
To my utter delight, in February, elements of two of my favorite programs are going to meld thematically as IBM's Watson supercomputer battles it out with history's most revered trivia champions, Ken Jennings and Brad Rutter.
The IBM researchers behind Watson are confident that they've achieved their goal to build a system that can rival mortal abilities to answer questions asked in normal speech--and to do so quickly. As IBM noted of the tournament, which will be held over three days beginning February 14th, this provides the "ultimate challenge because the game's clues involve analyzing subtle meaning, irony, riddles, and other complexities in which humans excel and computers traditionally do not."
IBM's champion has already sparred, with mixed success, against some of the brightest Jeopardy minds during a series of 50 sparring games in advance of the primetime debut, which can be viewed here (along with some neat history behind the machine and its capabilities). Upstairs, in a room filled with servers humming away, Watson was processing quietly…and giving human contenders a run for their money.
Following the airing of the trivia battle, there are very likely going to be fresh rounds of mainstream media comments on the man versus machine debate. News outlets will doubtlessly feature background images of Skynet and other Hollywood versions of "good machines gone bad" because, well, if Watson proves itself on national television, it's bound to be a little unnerving.
Instead of calling to mind deadly scenes from War Games and countless other sci-fi flicks that feature feral machines with minds of their own hell-bent on destroying their human makers, I think of a kinder, softer version of artificial intelligence (AI) -- Lieutenant Commander Data. After all, if there is any Hollywood counterpart to Watson (minus the emotion chip, thank goodness) it's pop culture's most-loved Android.
Watson is, in some ways, like a 21st-century prototype of Star Trek's Data. Both have the equivalent of millions of books built into a natural language processing-based algorithm, both are able to pick up on wordplay subtleties, making them, well, in some ways, "people-smart" and they both have the ability to kick some serious human tail in a head-to-head mental match.
What ordinary viewers might overlook is the power of this context awareness. Watson is not simply answering questions posed simply -- "he" is processing vast amounts of context-dependent information presented in natural human speech to arrive at an answer that is based on any number of factors. This is, in a word, absolutely groundbreaking. On a speech and context recognition front, in particular -- turning questions that require multifaceted layers of knowledge into answers that rely on a number of variables from any number of sources. The analytics involved are mind-boggling.
And isn't that what makes Data so fascinating?
The only real difference between Data and Watson, outside of hundreds of years of extra R&D for Data's wiring, is the human-like veneer -- both in terms of appearance and speaking ability. If there's one thing Watson lacks, it's the ability to sound like anyone (or anything) but HAL.
This type of artificial intelligence goes far beyond what most of us would consider the great question-answer machine -- the search engine. Research and development has been conducted for years to arrive at creating machines that can field simple verbal questions, but these have lacked the algorithmic complexity necessary to arrive at answers based on context.
Just as was at the heart of any number of scenarios involving Data, however, was the inherent question of whether or not, with any amount or development in AI there would be certain core levels of knowledge that even the sophisticated question-answer machines couldn't get to the heart of. Are there ways to trick to such a context-aware computer -- to "stump the chump?"
It will be interesting to see what questions Watson misses and look for patterns that might indicate why a machine might be less proficient at producing an answer with so much of the world's information storehoused.
But on another note, watching the humans lose to IBM's marvel of computer engineering would be almost as uncomfortable as watching Worf attempt yoga in a spandex Starfleet uniform. In all fairness to the mortal team, however, it did take IBM multiple tries to a beat a certain chess legend…
As a side issue, IBM's goal with Watson as a research proof of concept is staggering, but there is clear business value behind the move to take their act to television audiences. If IBM can position itself as a leader in the arena of context-aware machines, the sky is the limit. We already talk to and command our phones, cars, and anything else with a chip, using simple verbal prompts. The company that delivers the power of real speech recognition and processing to any array of consumer and business devices is set to grow.
At a recent conference on cybernetics, the president of the American Association for Artificial Intelligence, when asked what the ultimate goal was for the organization, replied "creating Star Trek's Mr. Data would be a historic feat of cybernetics, and right now it's very controversial in computer science whether it can be done. Maybe a self-aware computer can be put into a human-sized body and convinced to live sociable with us and our limitations… that's a long way head of our technology, but maybe not impossible."
While indeed, on the sociability, size, and cybernetics front Data is still just science fiction, the advancements that IBM is getting ready to put on display for the world in February seem like a positive sign when it comes to our ability to start interacting personably with our machines.
I'll be gearing up for the great battle on February 14th. And just to put this out there, if IBM engineers are able to hire Brent Spiner for a little extra context-aware voice-over work, if only for Watson's primetime television debut, there's a good chance my head will explode.
Posted by Nicole Hemsoth - January 13, 2011 @ 5:02 PM, Pacific Standard Time
Nicole Hemsoth is the managing editor of HPC in the Cloud and will discuss a range of overarching issues related to HPC-specific cloud topics in posts.
No Recent Blog Comments
10/30/2013 | Cray, DDN, Mellanox, NetApp, ScaleMP, Supermicro, Xyratex | Creating data is easy… the challenge is getting it to the right place to make use of it. This paper discusses fresh solutions that can directly increase I/O efficiency, and the applications of these solutions to current, and new technology infrastructures.
10/01/2013 | IBM | A new trend is developing in the HPC space that is also affecting enterprise computing productivity with the arrival of “ultra-dense” hyper-scale servers.
Ken Claffey, SVP and General Manager at Xyratex, presents ClusterStor at the Vendor Showdown at ISC13 in Leipzig, Germany.
Join HPCwire Editor Nicole Hemsoth and Dr. David Bader from Georgia Tech as they take center stage on opening night at Atlanta's first Big Data Kick Off Week, filmed in front of a live audience. Nicole and David look at the evolution of HPC, today's big data challenges, discuss real world solutions, and reveal their predictions. Exactly what does the future holds for HPC?