[Top] [All Lists]

Re: [ontolog-forum] IBM Watson on Jeopardy

To: "John F. Sowa" <sowa@xxxxxxxxxxx>
Cc: "[ontolog-forum]" <ontolog-forum@xxxxxxxxxxxxxxxx>
From: Ron Wheeler <rwheeler@xxxxxxxxxxxxxxxxxxxxx>
Date: Mon, 14 Feb 2011 00:05:15 -0500
Message-id: <4D58B80B.5000203@xxxxxxxxxxxxxxxxxxxxx>
On 13/02/2011 10:30 PM, John F. Sowa wrote:
> Ron,
> IBM used an earlier version of that technology to beat Kasparov
> back in 1997.
>>> It's difficult to find good metrics for evaluating AI systems.
>>> Chess was one, but it was too specialized.  Jeopardy is much
>>> closer to commonsense reasoning.
> RW
>> That much horsepower applied to chess would not even be
>> interesting let alone impressive.
> It was very interesting in 1997.  Various AI leaders had
> been predicting that a computer would beat the world chess
> champion in 10 years.  They predicted that every 10 years
> from the 1950s to the 1980s.  Eventually they were right,
> but it took a lot longer than anyone had expected.
> And no computer has yet been able to reach the master
> level in Go.  The rules of Go are simpler than chess,
> but the number of possible moves is much greater.
My feeling is that the amount of processing power (CPU's and memory) in 
Watson applied to Go would probably yield a winning player. We are now 
at a point where the tree pruning does not have to be very aggressive to 
achieve a reasonable time between moves.    (01)

The ability to compete with a human on questions of general knowledge 
and cultural facts with the typical Jeopardy questions that often 
require subtle reasoning which is occasionally beyond human 
comprehension is a much more serious challenge for a computer.
Would it get "What is 42?" as the question corresponding  to "The 
ultimate question of life and everything, the universe and everything."
if you were constructing the quiz, what would you use as a topic title 
so as not to give the whole thing away?    (02)

I suspect that the impact of Watson on Ontology as a field can be 
negative or positive.
It shows that you can get answers to questions using data that is much 
less structured than the traditional languages used for ontology research.
It raises the question about what is a "good enough" process. Does it 
have to always produce the right answer? Does it have to be a repeatable 
process that is subject to proof?    (03)

If a Watson was available on a "per Question" basis and only cost a few 
cents per answer, how many companies or governments would use it?    (04)

> John
>    (05)

Message Archives: http://ontolog.cim3.net/forum/ontolog-forum/  
Config Subscr: http://ontolog.cim3.net/mailman/listinfo/ontolog-forum/  
Unsubscribe: mailto:ontolog-forum-leave@xxxxxxxxxxxxxxxx
Shared Files: http://ontolog.cim3.net/file/
Community Wiki: http://ontolog.cim3.net/wiki/ 
To join: http://ontolog.cim3.net/cgi-bin/wiki.pl?WikiHomePage#nid1J
To Post: mailto:ontolog-forum@xxxxxxxxxxxxxxxx    (06)

<Prev in Thread] Current Thread [Next in Thread>