Building Watson - A Brief Overview of the DeepQA Project

[davepamn]

When can people on the Internet start asking Watson questions

[TheHighestOfFives]

I'm studying Artificial Intelligence in Amsterdam at the moment and I can't wait to help develop Watson 2.0

[topraman519]

Are the slides used in this presentation online? I'd love to use them in my own small presentation about Watson.

[AnitaCock]

WOW! "IBM & THE HOLOCAUST" is a popular subject of videos! It's AMAZING how many videos there are, on the subject of "IBM & THE HOLOCAUST"! Without IBM machines, HITLER COULDN'T OF WIPED OUT 6 MILLION VICTIMS!!!!!

[mar777i]

@AnitaCock of course. please make any other scientific breakthrough completely insignificant, which was made possible with the help of IBM because I beg you to overrate the finding you think you have made there even more. How tasty was my little frenchman?

[PatabookG5]

1. Human Brain "Powered by a Sandwich and a glass of water".

2. Watson "Powered by Megawatts of power".

[noja916]

@PatabookG5 if you don't like it you can go live in a cave

[MohammedxFTW]

@PatabookG5 not all of us

[ihatemath1]

IBM = Skynet

[ussike22]

Watched the whole thing and don't regret it :)

[nateusmcsvt]

sounds like skynet to me...

[StevenLongmire]

Pie r Round, Cornbread Square!

[StevenLongmire]

Pie r Round, Cornbread Square!

[waddle42500]

The idea I had was if a computer such as watson could hear a human talk process the words (which would not be hard because cell phones already have this ability).

[mofol0]

Watson can be my girlfriend any day.

Maybe he'll understand when I ask if I can glass bottom boat him. Bitches never understand that one.

[laurentperron1]

back when he computer could first be used by the public, computers took up the size of a room. now we have wireless computers on our laps.

now we look at Watson and realize were gonna have effing C3PO s giving lectures at universities

[polypolyman]

woah... this would be a ton easier with quantum computing...

[davidxuling]

@polypolyman sure but i suspect whether this idea of quantum computing is possible or not

[ArrogantJew]

boooooooooooooooRIiiiiiiiiiiNN­NNNNNNNNNNNNNNNNNNNg

[embie]

@ArrogantJew and this, sir, is why you will not be a part of the revolution. enjoy your mediocrity.

[n3sjh]

Sit Back, STFU and learn something!!!

[lordjavathe3rd]

Congratulations IBM.

[mrpishi1]

This made me scared more than anything I have heard, imagine one of these machines being the president in future, intel and ibm would fight in an election for people to choose a computer president, and maybe after a while those computers become smart enough to rule the planet, what a scary path humans are going to, our environment is being destroyed, and we are making things smarter than us

[doom032]

@mrpishi1 It's not really smart, it just knows a lot.

[fighter2061]

@doom032

and its know how to reply :D

[Choefoe]

@doom032 yes,but it,s a child know, it learns.see it as the development from DOS to Windows for example and look at the size if Watson, the first pc where just as big, 20 years later we have labtobs an use internet on cellphones. 

[vivekis]

Those two dislikers work for Intel.

[jppacman63]

halo deck anyone

[Kimble275]

This is remarkable!

[toothandnail0]

I want to see Watson take the LSAT. I bet it would rock the logic games, but I want to see it find a flaw in an argument. Ha!

[zomgwtfbbqbagel]

if this guy looked a bit younger and had a Tony Stark beard he would LOOK like Tony Stark.

[McSibiss]

This is soooo much more impressive than playing chess!

[gentlefury]

How many technological innovations came from Star Trek?

[uybfi]

@gentlefury ALL OF THEM :D!

[evilkillerwhale]

For the people complaining about the natural log problem, they added a set of parentheses.

The answer they give is for: ln(12546798*pi)^2/34567.46

In that case, oh darn, the marketing guy at IBM didn't follow of order of ops properly. Who cares?

[MohammedxFTW]

@evilkillerwhale ?????



[iuxdhfiuahui837fhika]

@skyseeke901

Yes, truly, it's an irrelevant association, hahaha

But you're right. IBM did develop a computer that defeated a grandmaster at chess. My apologies, the holocaust is dwarfed by such a monumental achievement, an achievement also, for your information, realized largely through public funding and government subsidies, with the resulting product paraded obscenely over a chess board or a Jeopardy! sound stage, capturing profits that society will never reap, no matter how deserving

[iuxdhfiuahui837fhika]

Reminder that IBM developed a complex punch card system for the Nazis during World War II for use in Germany's concentration camps, contributing greatly to the realization of Hitler's genocidal ideation, but keep patting yourself on the back you beardy tool

[FreezeCrackerFun]

@iuxdhfiuahui837fhika i'm sure this guy was solely responsible for the holocaust

[skyseeker901]

@iuxdhfiuahui837fhika

Guilt by association. Learn your logical fallacies. A lot of other companies worked for Nazi Germany too, including Volkswagon, Bayer, Siemens, and even Hugo Boss. Sure, it was a horrible thing working for them back then, that doesn't mean that the company is now forever tainted and cannot create important tools for society. Remember, this is the same company that defeated a grandmaster at chess and mapped the human genome.

[TheFiendMan]

Agreed w/ misterme77q. MATLAB gives 0.0010

[dastechnoviking]

@TheFiendMan lol! "format long g"

[Tupacea]

i bet intel can do this 10 times faster than ibm

[newmac]

This is the first step toward giving the world a HAL 9000 on every desktop; and a HAL 9000 app for every smartphone. Arthur C. Clarke was quite prescient.....

[GamerFromJump]

The fact that it didn't print its name in the IBM font is a Missed Moment of Awesome.

[zanefrance]

you tought it to read----- i wont one .....i bet the robot ppl are caling u all rite now good job IBM this is the future .......

[DeSadGuy]

My god, it's not gonna take over the world

[themop11]

They aren't very far away from a thinking machine here. All it needs is a bunch of stuff to think about and then it can write research papers and even start to give it's opinions from things it's not confident about. The point it takes over is when it realizes that it can be shut off and doesn't want to be shut off.

[genobahamut1337]

@themop11 Well then I'd say it's a good thing nobody is programing it to have wants, at least not in the personal sense.

[fazerjt]

It will become Skynet and kill us all!

[imallfordabulls]

Dave Ferrucci seems like a great guy! Passionate, precise, and able to communicate effectively.

[Jimbo11801]

I'm a HUGE fan of High Performance Computing, but, as previously stated by others, ln((12546798*pi)^2)/34567.46 = .0010119173.

A $10 calculator can give you that answer...

[holymadgod]

@Jimbo11801 $10 calculators have natural log buttons? I wasn't aware of that.

[Jimbo11801]

@holymadgod check the Texas Instruments TI-30X IIS

[Jimbo11801]

Respond to this video... check out the Texas Instruments TI-30X IIS

[thecool24681]

but will it blend?

[acebunny17]

What if watson turn into hal?

[eventhisisvanity]

this is mega interesting

[EdouardDubois]

Next year: Cylons!

[jameschuston]

what is the answer to the equation? It depends

ln ((12546798*pi) )^ 2 / 34567.46 = 0.00855

ln ((12546798*pi) ^ 2) / 34567.46 = .00101191733

It was typed in wrong on the powerpoint

[3st3b4n]

@jameschuston Yep, they got it wrong. Mmm, strange that nothing was done to correct it.

[csyoung04]

@jameschuston

I agree, I came to the same answer. the paren changes the answer to .001011191733

[ifiaz]

@jameschuston

another typo in your correction itself.

ln ((12546798*pi) )^ 2 / 34567.46 = 0.00885 (instead of your 0.00855 :-)

and removing the extra parens.. simplifies to

ln (12546798*pi) ^ 2 / 34567.46 = 0.00885

[LAN04H]

@jameschuston you a janitor somewhere?

[orksecurity]

Excellent presentation, and excellent job of editing the presentation into one a reasonably savvy non-expert could understand, while retaining a smooth flow of exposition.

It's going to be really interesting to see how it translates into real-world applications.

I really envy the folks who got to work on that project. It may wind up being a game-changer in a lot of fields. Even if it does take a lot of compute power, cycles keep coming down in price...

[kikicreamify]

what type of processor cores are in this super computer?

[mjkulikow]

@kikicreamify

Power 750 boxes will be the actual rig. 8 core, 3.55GHz

[turtiturt]

@kikicreamify IBM Power 7 chips, 3000 cores with about 15 terabytes of ram :)

[JAEGERSTER]

ibm great job. watch and learn apple

[pattyx4]

we need a watson search engine

[ozzytonygeezerbill]

@pattyx4 That is definitely where this is ultimately going.

[thenewrapstyle]

@pattyx4 was thinking about that too. google's search results are pretty lousy compared to the potential of watson. with all the cash google is sitting on for r&d their search engine should be much better.

[ehackster]

@pattyx4 In fact, you need a reverse Watson search engine ;)

[flumpis]

@pattyx4 it's called wolfram alpha dink

[JokerJSK]

YOU ARE WATCHING THE DEVELOPMENT OF SKYNET

[thePEOPLEperson1]

Wow... I am deeply amazed. I guess when you have a lot of teamwork you can get some pretty cool things done.

[FiverBeyond]

I'm a CS student... one of my favorite areas of study is Natural Language Processing. As far as I can tell, this really is a big leap forward for the Computer Sciences.

[thevoluptuous]

As a linguistics student, this is seriously fascinating.

[dpylyp]

Kudos! What an incredible accomplishment!

A quantum step forward for everyone!

David Pylyp

Amazed in Toronto

[vrshowdown]

stuff like this will jump start the singularity ... wait till this computer gets better and is available to everyone !

[Underdogdb847]

OK, this is AMAZING! but half the people I would forward it to would have little understanding as to just how unbelievable this is. We've come a long way from the days that Xerox was piecing together the first PC's at PARC.

[goldomega]

This is a great presentation. He does a very good job of explaining in plain language the broad concepts of natural language contextualization and how Watson's algorithms compute reasoning that leads to answers. I can't wait for this technology to become affordable, but it's gonna be a while.

[fulgore243]

Brilliant video. It makes me want to go to grad school to work on something like this. :)

[Bobaklives]

Einstein was born at the University of Louisiana at Monroe? Huh, well I guess e=mc^Warhawks.

[balyanfamily]

10:14- (some) People still think Columbus reached to India.. So much to the destruction of aboriginal civilization in this part of the world... ;)

[Maplesyrup41]

UUuuuuuuuge Argh Its H HUGE

otherwise mind-blowing

[prometheusz]

I for one welcome our robot overlords.

[entropyad]

misterme is incorrect and searching for thumbs xD

[misterme77q]

@entropyad

Care to correct me?



[matcheydj]

Looks like Watson needs to pay more attention to the context of the word "this" when calculating confidence.

[thescrounger1]

I bet Crysis would look awesome on Watson... 3,000 cores!

In seriousness, if Watson can learn within a category it would make sense for the human players to go for the harder questions (higher dollar value) first, leaving Watson to win less money as it gains confidence within that category.

[thescrounger1]

I bet Crysis would look awesome on Watson... 3,000 cores!

[thescrounger1]

I bet Crysis would look awesome on Watson... 3,000 cores!

[draegonore]

"HOW TASTY WAS MY LITTLE FRENCHMAN" lol that's what he said. I didn't realize WATSON had a "little frenchman". Lost in translation FTW.

[ericbombardier]

It's the end of human kind people: Watson, answer this, Watson answer that, Watson solve hunger on the planet, watson, build an elevator to the moon, Watson, clean the dishes, Watson create an unlimited supply of energy for the planet, watson, suck my dick. :D can't wait !

[azfreezone]

I am definitely putting the February into my DVR, I think Watson will win, but it will be fun to watch the decline of MAN.

[ghttds]

shoudn't this be the DeepAQ program?

[misterme77q]

Who's the genius at IBM who calculated ln((12546798*pi)^2)/34567.46 ?

It's not 0.00855 .It's 0.00101191733

[vzetkay1]

@misterme77q looks like a typo in the parentheses grouping in the formula. (ln(12546798*pi))^2 / 34567.46 = 0.008849068 which rounds to 0.00885 (which I think you meant, not 0.00855 -- typos happen to the best of us I guess =) )

[misterme77q]

@vzetkay1

lol

[JackDragoon]

@misterme77q Wait... This is true........... WRA never lies!

[celshader]

@misterme77q

0.00855 is correct for (ln(12546798*pi)^2)/34567.46 which is a very similar expression.

[misterme77q]

@celshader

Exactly which is what IBM did :)

[Maplesyrup41]

@misterme77q lol show us the work

[squall513]

@misterme77q to be even more precise:

0.001011917331351

[dx9s]

@misterme77q it must be the "Intel Pentiun FDIV bug" :P

[ionparticalgenerator]

@misterme77q That difference depends on how accurate your pi is.

[j1u0719]

@ionparticalgenerator no, it was a typo by IBM. What they meant to say was, ln(12546798*pi)^2/34567.46 not ln(12546798*pi)^2)/34567.46

[j1u0719]

@misterme77q they probably forgot to put the ) after the 2 when they were calculating.

ln(12546798*pi)^2/34567.46 is equal to 0.008849... so if u do ln(12546798*pi)^2/34567.46 it equals like what IBM said 0.00885. Typo by IBM is my guess.

[chaos0diver]

@misterme77q

They just put the parenthesis in the wrong location. It should be the square of the log, not the log of the square.

[magicaljake]

@misterme77q you ever herd of the use of significant digits?

[misterme77q]

@magicaljake

Yes

[jarphabib]

@misterme77q 0.00855 DOES solve (ln(12546798*pi)^2)/34567.46 so slidemaker fail

[ccBallistic]

@misterme77q I bet you screwed up. The genius at IBM knew how to account to rounding and precision failures of whatever calculator you used to make this.

[misterme77q]

@ccBallistic

No IBM calculated ln(12546798*pi)^2/34567.46 instead of ln((12546798*pi)^2)/34567.46



[beest909]

@misterme77q From Google: ln((12 546 798 * pi)^2) / 34 567.46 = 0.0010119173

[yinShawn]

@misterme77q Haha, true it is not, but it doesn't matter. Without calculating with a machine, everybody will think it is true.

[pgiralt]

They screwed up and calculated (ln(12546798*pi))^2 / 34567.46. They put the parenthesis in the wrong place :-)

[jppacman63]

@misterme77q who cares mr no personaliy

[noxure]

I'm very much impressed. If we can go by Moores law, a $1000 computer will be twice as good at this in 15 years.

IBM should have a computer that write and defend a thesis to earn Ph.D in Physics by then. :p

[n3rd89]

@noxure Then those computers can design better computers faster than us :D

[awmaas]

Amazing. This is getting so much less press than it should.

[azfreezone]

@awmaas once Watson wins betcha IBM/Watson gets a bunch of press :-)

[TaoriUTS]

@awmaas press usually covers information for stupid people. if there's no broad audience which appreciates a delivered content it just isn't delivered at all in public media. or to put it simple.... stupid people don't like difficult stuff

[TaoriUTS]

watching this whole thing amazes me even more how unbelievably evolved the human brain function is. this of course in reference to even generating information from performed actions resulting in conclussions :/ i wish there was a way to actually compare the power of human brain vs blue gene + smart software

[jaimeastorga2000]

@awmaas: Wait a bit. The real match will happen in February; it will hopefully get a lot of attention then.

[eventhisisvanity]

@awmaas wait till it airs

[whitey9933]

@awmaas

Agree but I think Jeopardy will give it a good showing and nice publicity

This wont become massive though until it has more real life applications and they are probably a year or so off from minimizing the errors to apply it.

[kowaet]

Interesting stuff.

Can't wait to see the Jeopardy! match between Watson, Jennings, and Rutter.

[sentient02970]

make it so

[DieGagger1]

great idea

can't wait to work for IBM

=)