IBM Watson Supercomputer

amirm

Banned
Apr 2, 2010
15,813
38
0
Seattle, WA
Supercomputers had a godly status in 1970s and 1980s. Then came the era of fast PCs and parallel arrays of them and the era came to an end. While massively parallel computers are put to use, they have lost the luster they used to have in the eyes of everyday man.

IBM seems to be on the path to bring back some of that glory through combination of massively parallel supercomputers *and* software in the form of "Watson." This article nicely outlines its public accomplishments. It shows what you can do with a system that has a ton of storage, computing power and some smarts in the form of software:

http://www.dailytech.com/article.aspx?newsid=23475

-----

The Strategic IP Insight Platform provides the perfect tool for finding prior art -- or potential victims

International Business Machines, Inc. (IBM) sets a high bar for future artificial intelligence machines. With its powerful parallel software package DeepQA -- written in a mix of C++, Java, and Prolog -- IBM's Watson supercomputer learns, data mines, performs image analysis, processes natural language, and understands speech in a unified approach that eclipses the more humble machines that came before it.

Despite a very human flub, Watson put its powers on display on the game show Jeopardy, accruing $77,147 USD for the win in a match against the longest time Jeopardy champion -- Ken Jennings (who scored $24,000) -- and the biggest money winner on the show -- Brad Rutter (who scored $21,600). The humans' loss led Ken Jennings to quip, "I for one welcome our new computer overlords."

I. Watson Enters the Patent Industry, Scouring Medical Patents

Now Watson is applying that same cold, masterful logic to the patent industry. Dubbed the Strategic IP Insight Platform (SIIP), Watson's new software allows it to delve peer-reviewed literature and past patents and build a rich tapestry of what has been done before. At the same time it's cataloging various firms, analyzing their intellectual property, products, and financials.

Ideally the patent industry is about coming up with truly novel, non-obvious solution, and seeing that solution protected so you can profit off your ingenuity. While the concept of an "obvious" solution is the kind of sticky philosophical debate that may still be beyond the mighty AI's grasp, Watson can certainly tackle whether a patent is novel -- i.e. whether there is prior art that should overrule it.

Top Jeopardy player Ken Jennings acknowledge Watson as his superior on national television.


In a demonstration of its power, Watson has 4.7 million patents and 11 million scientific journals published between 1976 and 2000. In each article Watson carefully identified any previously unseen chemical compounds (in pictures or text), grabbed any related diagrams, grabbed relevant keywords, and lastly scooped up the author and company names.

The result was a database of 2.5 million compounds, which are thought by Watson to be unique. For each of these compounds Watson discovered the earliest patentee. IBM donated its superbot's work -- the "open chemistry" database -- to the U.S. National Institute of Health (NIH), allowing scientists all over the world to dig into it.

Watson has identified 2.5 million unique medical compounds.

The work means that Watson has now implanted itself in both in the pharmaceutical industry, and the medical diagnostics business. Previously, Watson teamed up with Wellpoint, Inc. (WLP) to given a patient's medical history and symptoms, to scour medical literature and provide probabilistic differential diagnoses of known maladies (as seen in series of ads).

II. Watson Could be Used as a "Nuclear" Weapon by Innovators and Trolls, Alike

Returning to SIIP, the U.S. Patent and Trademark Office (USPTO) and other large international patent offices might be wise to recruit Watson's expertise. Sadly it is not uncommon these days for people to file, be granted, and even sue with patents on an idea that had already previously been patented by someone else. Using Watson, the USPTO and its peers could cut down on redundant patents, protecting the original inventor.

But in the long term, the greater questions is whether Watson's capabilities will be applied for good or for evil.

Watson's SIIP can set to work scrounging up prior art, giving companies a powerful defense against patent trolls, saving potentially millions in fees to patent lawyers or experts whose findings would likely be less complete. And SIIP also has the capability to assess which topics to target with research, assess undervalued companies with attractive IP worth acquiring, or similarly companies that could be valuable collaborators.

Those powerful abilities could potentially fall into the wrong hands and be used for "evil". A patent troll (e.g. NTP or ex-Microsoft Corp. (MSFT) CTO Nathan Myhrvold's Intellectual Ventures) could use Watson's data mining to assess companies' IP and financials, compiling a list of targets that would likely fall easy prey to a lawsuit and be forced to settle or license.

Watson's IP prowess could be a powerful weapon for a patent troll looking to profit at the expense of innovation.

The battle between good and evil is in fact one that appears to go on at IBM itself. With over 50,000 patents, IBM has 25 percent more patents than its next closest international rival -- Samsung Electronics Comp., Ltd. (KS:005930). Many of those patents are novel in the very best sense possible. Others, like IBM's patent on ignoring Lotus Notes seem pretty questionable. Indeed, it's probably expected that IBM -- who filed for over 6,000 patents in 2010 -- would be a microcosm of the greater patent industry.

III. Next Generation Watson to "Think" -- But Know More Than Any Human Can

Looking ahead, IBM could see its products cast even deeper into the battle between "good" and "evil" in the patent industry. While Watson cannot "think" like a human, IBM is working to make sure its successor can. The company is developing neurosynaptic chips, which it looks to deploy in a cognitive-capable exascale (beyond petascale) computer within the next decade.

If you think Watson's current generation capabilities are impressive, its successor will be granted cognitive abilities, which will give it the ability to not only "out-know" its fleshy rivals, but "out-think" them as well.

For hardware fans out there, Watson packs ninety Power 750 four-socket server nodes, with a 3.5 GHz POWER7 octacore processor in each node. That's a total of 2,880 CPUs (with four threads each, so 11,520 total threads).

Paired to that is 16 TB of memory (about 182 GB per server) of RDIMM (registered DIMM, lower latency) memory and your typical networking, cluster, and I/O support nodes.

In order to win Jeopardy Watson only needed a modest 4 TB, which it stored its 200 million pages of relevant information in.

In related news, Watson has also gone to work selling its own capabilities and other IBM products, acting as a "smart" telemarketer. During sales efforts, IBM is now using Watson as a qualified associate to answer tough questions customers might have -- a problem for under-qualified associates at your average business.

IBM recently scored legendary investor Warren Buffet, typically a staunch critic of tech firms, as an $10.7B USD major investor. Perhaps Mr. Buffet was swayed by IBM's strong track record -- currently in its 101st year of business, IBM is the oldest of the large technology companies.
Source: SIIP
 

flez007

Member Sponsor
Aug 31, 2010
2,915
36
435
Mexico City
Amir - I participated in a huge project here with the University of Mexico where a +20,000 parallel Intel CPU "supercomputer" was deployed for animation, high transactional and simulation research projects - even the fastest multi-core UNIX server felt behind for complex mathematical functions compared to this blade server, and it is stackable!
 

amirm

Banned
Apr 2, 2010
15,813
38
0
Seattle, WA
I remember days where people scratched their heads trying to get scaling to 8 and 16 CPUs! Now we have tens of thousand :).

Whey did they buy that system?
 

Orb

New Member
Sep 8, 2010
3,010
2
0
Ah I miss the days of the monster water cooled IBM mainframes, for me data centre rooms seemed a bit more Star Trekkie in those days :)
Last one I ever worked on was the 1st installation of a special watercooled ES/9000, discounted down to the price of £10mill (back in 1990) as it had never been built outside of an IBM lab :)
Now those were fun days taking out and putting in a totally new (and at time one of the more powerful ones existing) mainframe within 36 hours (with all OS and applications running again)..

Just seemed more special to me back then, or I just got older :)
Bit envious of those who get to work with the modern true research mainframes, and associated connected hardware.

Cheers
Orb
 

flez007

Member Sponsor
Aug 31, 2010
2,915
36
435
Mexico City
I remember days where people scratched their heads trying to get scaling to 8 and 16 CPUs! Now we have tens of thousand :).

Whey did they buy that system?

The rationale was R&D purposes, particullary for 3D imaging of complex functions related to bio-tech models, but other practical implementations such as batch processing for balance consolidations at FSIs are now being consider.
 

flez007

Member Sponsor
Aug 31, 2010
2,915
36
435
Mexico City
Ah I miss the days of the monster water cooled IBM mainframes, for me data centre rooms seemed a bit more Star Trekkie in those days :)
Last one I ever worked on was the 1st installation of a special watercooled ES/9000, discounted down to the price of £10mill (back in 1990) as it had never been built outside of an IBM lab :)
Now those were fun days taking out and putting in a totally new (and at time one of the more powerful ones existing) mainframe within 36 hours (with all OS and applications running again)..

Just seemed more special to me back then, or I just got older :)
Bit envious of those who get to work with the modern true research mainframes, and associated connected hardware.

Cheers
Orb

We are heading back there Orb, new SaaS and Cloud Computing models rely on very powerfull Data Centers serving mobile and thin clients worldwide.
 

amirm

Banned
Apr 2, 2010
15,813
38
0
Seattle, WA
The rationale was R&D purposes, particullary for 3D imaging of complex functions related to bio-tech models, but other practical implementations such as batch processing for balance consolidations at FSIs are now being consider.
Sorry, I meant to say when :).
 

amirm

Banned
Apr 2, 2010
15,813
38
0
Seattle, WA
Ah, that's less extreme :). The IBM system has an advantage as the Power cores are smaller than Intel's X86 so cost less to make more of them.
 

Ronm1

Member Sponsor
Feb 21, 2011
1,745
4
0
wtOMitMutb NH
Ah I miss the days of the monster water cooled IBM mainframes, for me data centre rooms seemed a bit more Star Trekkie in those days :)
Last one I ever worked on was the 1st installation of a special watercooled ES/9000, discounted down to the price of £10mill (back in 1990) as it had never been built outside of an IBM lab :)
Now those were fun days taking out and putting in a totally new (and at time one of the more powerful ones existing) mainframe within 36 hours (with all OS and applications running again)..

Just seemed more special to me back then, or I just got older :)
Bit envious of those who get to work with the modern true research mainframes, and associated connected hardware.

Cheers
Orb

I know how you feel, was at CDC for a few years. H/W was quite impressive, especially the STAR, consider R&D wise it was picked over the 8000 which went on to become the Cray1. Bad decision biz wise as Seymour was right, but h/w was impressive nonetheless.
 

flez007

Member Sponsor
Aug 31, 2010
2,915
36
435
Mexico City
Ah, that's less extreme :). The IBM system has an advantage as the Power cores are smaller than Intel's X86 so cost less to make more of them.

This project ramped-up via donations from the private sector, we gave some SYbase ASE licenses for DB replication, what I was told from UNAM officials is that the configuration could easily grow to 2x or 3x with no technical issues except available space, cooling systems and of course additional SW License fees - this is a very interesting and promising project to simulate astronomic models simulations.
 

About us

  • What’s Best Forum is THE forum for high end audio, product reviews, advice and sharing experiences on the best of everything else. This is THE place where audiophiles and audio companies discuss vintage, contemporary and new audio products, music servers, music streamers, computer audio, digital-to-analog converters, turntables, phono stages, cartridges, reel-to-reel tape machines, speakers, headphones and tube and solid-state amplification. Founded in 2010 What’s Best Forum invites intelligent and courteous people of all interests and backgrounds to describe and discuss the best of everything. From beginners to life-long hobbyists to industry professionals, we enjoy learning about new things and meeting new people, and participating in spirited debates.

Quick Navigation

User Menu

Steve Williams
Site Founder | Site Owner | Administrator
Ron Resnick
Site Co-Owner | Administrator
Julian (The Fixer)
Website Build | Marketing Managersing