Jeopardy: Human vs. Machine

diogen

SatelliteGuys Pro
Original poster
Apr 16, 2007
4,313
0
Interesting details
10 racks full of IBM Power 750 servers, stuffed with 15 terabytes of RAM and 2,880 processor operating at a collective 80 teraflops. IBM says it would take one CPU over two hours to answer a typical question...
So, assuming 100% parallelilism (2 CPUs do the job twice as fast, 10 CPUs - 10-times as fast, etc.) this monster will have an answer in 2*60*60/2880, i.e. 2.5 seconds.
http://www.engadget.com/2011/01/13/ibm-demonstrates-watson-supercomputer-in-jeopardy-practice-match/

Regardless of this competition outcome, it is amazing what it takes to match some capabilities of about 1 liter material called brain.
That some claim isn't even used 100% and isn't digital to boot...

Diogen.
 
Especially liked this response:
Q: Humans, do you feel any pressure competing with Watson? A: Brad isn't worried about Watson, but afraid of "Watson's progeny when they come back from the future to kill me." Ken says he was warned by a friend, "Remember John Henry.' Ken's response: "Screw that, remember John Connor!"
 
When we start talking about few seconds to find an answer to a question (any question; and every question in
Jeopardy has an answer!), the prediction algorithm will be at least as important as number crunching abilities.

If the format doesn't change, the question is read out aloud while shown on the screen at the same time.
Picking it up from the screen should be a millisecond job for the machine, i.e. it will have a couple second advantage (while it is read).
The rest will be database access times and cache efficiency...

The interesting question is whether the machine will have a "guessing" algorithm,
i.e. will it attempt to guess an answer when the "standard" search fails.

Diogen.
 
When we start talking about few seconds to find an answer to a question (any question; and every question in
Jeopardy has an answer!), the prediction algorithm will be at least as important as number crunching abilities.

If the format doesn't change, the question is read out aloud while shown on the screen at the same time.
Picking it up from the screen should be a millisecond job for the machine, i.e. it will have a couple second advantage (while it is read).
The rest will be database access times and cache efficiency...

The interesting question is whether the machine will have a "guessing" algorithm,
i.e. will it attempt to guess an answer when the "standard" search fails.

Diogen.

I think you meant that every answer in Jeopardy has a question!
 
I think you meant that every answer in Jeopardy has a question!
Hehe... Not really.
Every answer anywhere has at least a few questions it answers.
But not every question in life has an answer. In Jeopardy - it does!

Diogen.
 
With Moore's law staying strong for decades, we have achieved more than we thought we will over the last 50 years.
Without computers, i.e. number crunching capabilities well beyond a human reach, it is really hard to imagine what out race would be today.

But when compared to a human brain, it looks like we are building a Caterpillar with hopes to beat a Porsche.
It certainly will in many things. But not in all...

Diogen.
 
The one thing that I found funny is that IBM did not even use their own OS (AIX) for this system. They resorted to using SuSE Linux for the OS over AIX.
 
I was very surprised as well. I figured it was AIX when I heard about the project. My Novell account manager sent me a little better information on the whole project when they released the information to the media.
 

Users Who Are Viewing This Thread (Total: 0, Members: 0, Guests: 0)

Who Read This Thread (Total Members: 1)

Latest posts