IBM's Watson Has Gone to College
Even supercomputers need a little schooling.
Seemingly paying homage to Star Trek's Data, Intel has reportedly sent its Watson supercomputer to college so that it can learn to be "more human".
According to IBM, the company is actually sending a modified version of Watson to the Rensselaer Polytechnic Institute, marking the first time a Watson system will be used in a university. Watson was made famous by nuking the world's best human players on "Jeopardy!" in 2011, but now it's heading to college to sharpen its mathematical ability, to figure out how to quickly understand the meaning of new or made-up words, and more.
"The leadership of Rensselaer faculty in Web science, Big Data, artificial intelligence, and other research areas uniquely situates the university to help expand Watson’s abilities," IBM said. "Rensselaer faculty and students will seek to further sharpen Watson’s reasoning and cognitive abilities, while broadening the volume, types, and sources of data Watson can draw upon to answer questions."
The Watson system provided to the university will have 15 TB of hard disk storage, allowing it to store roughly the same amount of data as the "Jeopardy!" model. It will also allow up to 20 users to access it simultaneously, thus creating an "innovation hub". In addition to faculty researchers and graduate students, undergraduate students will also have access to help prepare them for future high-impact, high-value careers in analytics, cognitive computing, and related fields.
IBM said that Watson has a deep connection with Rensselaer, the oldest technological university in the United States, as many graduates became key members of IBM's Watson project team. "Leading up to Watson's victory on Jeopardy!, Rensselaer was one of eight universities that worked with IBM in 2011 on the development of open architecture that enabled researchers to collaborate on the underlying QA capabilities that help to power Watson," IBM added.
The company said it will provide Rensselaer with Watson hardware, software and training.
In addition to Rensselaer, IBM is also collaborating with medical providers, hospitals and physicians via Watson to help doctors analyze a patient's history, symptoms and the latest news and medical literature, thus speeding up diagnoses and making them more accurate. IBM is also working with financial institutions to help improve and simplify the banking experience.
Stay On the Cutting Edge: Get the Tom's Hardware Newsletter
Get Tom's Hardware's best news and in-depth reviews, straight to your inbox.
-
dameon51 I wonder how long until this tech is phone sized. Any of you Moores Law/math whizzes wanna crunch some numbers?Reply -
Northwestern dameon51I wonder how long until this tech is phone sized. Any of you Moores Law/math whizzes wanna crunch some numbers?20 years tops.Reply -
Northwestern20 years tops.Sounds accurate but don't forget that they'll also find ways to make it more efficient, shave some features off for the mobile version, etc.Reply
-
ddpruitt dameon51I wonder how long until this tech is phone sized. Any of you Moores Law/math whizzes wanna crunch some numbers?Reply
The problem isn't shrinking the technology down, it's getting the technology to the point anyone can use it. They Watson (and related systems) work is that the questions have to be properly formatted. Although Watson cleaned up in Jeopardy the questions are worded in a consistent manner and the machine was tweaked to be able to handle the questions. Even supercomputers have problems with natural language processing. Just look at Siri and it's equivalents, every request is sent to a server farm and even then it needs a properly formatted query.
We can build smart computers, but we've been working to build computers that can understand people for half a century. I figure we need another 15 or so years to have a big computer understand people, then another 7 or 8 to make it available on home computers (desktops/laptops) and another 7 or 8 for low power mobile devices. -
sacre I'm impressed, but at the same time not so much. I would like it to be able to scan data on its own and try and solve complex problems without our help.Reply
Basically like how we try and figure out what Dark matter really is, we'll tell it about dark matter and let it try and come up with ideas or discoveries by itself with it scanning and thinking.
But unfortunately we're far from that because it requires programming.. and tons of it.. For a machine to discover things by itself.. Wanting to discover, driven by curiosity?
Thats when i'll be even more impressed. -
sacre ddpruittThe problem isn't shrinking the technology down, it's getting the technology to the point anyone can use it. They Watson (and related systems) work is that the questions have to be properly formatted. Although Watson cleaned up in Jeopardy the questions are worded in a consistent manner and the machine was tweaked to be able to handle the questions. Even supercomputers have problems with natural language processing. Just look at Siri and it's equivalents, every request is sent to a server farm and even then it needs a properly formatted query.We can build smart computers, but we've been working to build computers that can understand people for half a century. I figure we need another 15 or so years to have a big computer understand people, then another 7 or 8 to make it available on home computers (desktops/laptops) and another 7 or 8 for low power mobile devices.Reply
Will it truly ever "understand" what is said or will it just processed based on its input and supply a proper response?
A machine would have to be aware to truly understand and comprehend what is said to it. Could that be achieved? Well, what is life? Where does it go from random movements in a molecule to life? Is life nothing more than a complex biological system? If so why can a complex mechanical system have life?
We have a basic programming in us just as a machine does, except ours can grow and expand on its own whereas machine today is extremely limited to what it can "learn" and how it can utilize that information.
-
rebel1280 sacreWill it truly ever "understand" what is said or will it just processed based on its input and supply a proper response?A machine would have to be aware to truly understand and comprehend what is said to it. Could that be achieved? Well, what is life? Where does it go from random movements in a molecule to life? Is life nothing more than a complex biological system? If so why can a complex mechanical system have life? We have a basic programming in us just as a machine does, except ours can grow and expand on its own whereas machine today is extremely limited to what it can "learn" and how it can utilize that information.YOUR A CYLON!Reply
-
linkgx1 I wonder when Cyberdyne will be possible ! :)Reply
I already have my lawsuit documents prepared against IBM! -
aramisathei sacreI'm impressed, but at the same time not so much. I would like it to be able to scan data on its own and try and solve complex problems without our help. Basically like how we try and figure out what Dark matter really is, we'll tell it about dark matter and let it try and come up with ideas or discoveries by itself with it scanning and thinking.But unfortunately we're far from that because it requires programming.. and tons of it.. For a machine to discover things by itself.. Wanting to discover, driven by curiosity? Thats when i'll be even more impressed.Reply
Isn't that how "The Matrix" started? -
shadowfamicom Only 15TB? It may be only text based data (not sure) but I expected way more then that. Even my main rig and smaller one have 15tb in 3tb drives when combined. (if you must know, 3D models, media and archived data going back to 2006). If you had told me Watson used closed to a petabyte I would have believed you.Reply