Tech commentator - "Jeopardy" competition was unfair
David
*
*
*
*
*
*
*
*
http://www.tvweek.com/blogs/tvbizwire/2011/02/why-former-jeopardy-champ-ken.php
Tech Commentator: Watson Won on 'Jeopardy!' Because the Exhibition
'was set up in a blatantly unfair way to favor the computer.' Also, He
Says About IBM's Watson--Which Trounced Its Human Opponents--'All very
rudimentary work for a computer'
While many of us may be impressed at the trouncing IBM's supercomputer
Watson gave to "Jeopardy!" champs Ken Jennings and Brad Rutter this
week, the director of the Government Computer News (GCN) labs was
unimpressed.
Writing today, Feb. 17, 2011 on GCN's website, John Breeden II said,
"You would think that a techie like me would be overjoyed at seeing a
glimpse into the rise [of] what show loser Ken Jennings called
[referring to a 'Simpsons' episode] 'our new computer overlords.' But
actually I was disgusted by the whole dog and pony show, which was set
up in a blatantly unfair way to favor the computer."
That's because, Breeden writes, "Watson had an unfair advantage: It
was being fed the questions electronically. I wanted to see Watson
hear the questions using speech recognition and process them the way
humans do. But Watson was instead fed the words that made up the
question in ASCII text and then went about searching a database,
albeit a good one, looking for patterns and coming up with the proper
response. All very rudimentary work for a computer, actually, and not
much different than what Google and Bing do every day right now."
Jennings and Rutter played Watson for two games. After the final game,
televised on Wednesday night, Feb. 16, 2011, the tally was $77,147 for
Watson, $24,000 for Ken Jennings and $21,600 for Brad Rutter.
According to a story in The New York Times, IBM researchers
acknowledged that Watson benefited from the "buzzer factor," because
its human competitors had to anticipate the light that signals it's
possible to "buzz in" to provide a response, while the computer uses a
weighted scheme that allows it to buzz in as quickly as 10
milliseconds. That made it hard for the human competitors to hit their
buzzers in time.
Furthermore, Breeden writes, "The fact that Watson had a buzzing
device is irrelevant. It already knew how it would answer before the
question was finished being read, and the humans were still gathering
input. And considering that the questions on last night’s show were
actually pretty easy for “Jeopardy!” and that Jennings and Rutter
obviously knew most of the answers, what Watson really excelled at was
buzzing in faster than the humans."
We here at TVWeek noted that sometimes, when a answer was written in a
way more easily understood by humans, Watson had trouble. For example,
the Final Jeopardy answer for game one, in the category U.S. Cities,
was: "Its largest airport is named for a World War II hero; its second
largest, for a World War II battle." Both Jennings and Rutter got the
questions right--What is Chicago (which has O'Hare and Midway
airports). Watson was completely lost and came up with "What is
Toronto."
The WGBH PBS science show "Nova"aired an excellent program this month
about IBM's Watson project, and it's well worth watching:
Patty Winter
In article <n8trl65juge2jpjbl...@4ax.com>,
David <diml...@yahoo.com> wrote:
>
>http://www.tvweek.com/blogs/tvbizwire/2011/02/why-former-jeopardy-champ-ken.php
>
>Furthermore, Breeden writes, "The fact that Watson had a buzzing
>device is irrelevant. It already knew how it would answer before the
>question was finished being read, and the humans were still gathering
>input.
Uh, wha? Does he think that Rutter and Jennings can't read faster
than Alex Trebek can talk?
Patty
tdciago
> Uh, wha? Does he think that Rutter and Jennings can't read faster
> than Alex Trebek can talk?
Another person who may know a whole lot about technology, but precious
little about playing "Jeopardy!"
He also clearly has no idea that anticipating the buzzer is what
successful contestants *always* do.
But we've addressed all this stuff ad nauseum the last few days,
haven't we?
:)
Ken Wesson
> Writing today, Feb. 17, 2011 on GCN's website, John Breeden II said,
> "You would think that a techie like me would be overjoyed at seeing a
> glimpse into the rise [of] what show loser Ken Jennings
It says a lot about the state of our economy that someone who won
$300,000 is considered to be a "loser". :)
> Furthermore, Breeden writes, "The fact that Watson had a buzzing device
> is irrelevant. It already knew how it would answer before the question
> was finished being read, and the humans were still gathering input.
Poppycock. Everyone reads the clue off the screen before Alex is done
reading it.
> The WGBH PBS science show "Nova"aired an excellent program this month
> about IBM's Watson project, and it's well worth watching:
That colon makes it look like there's supposed to be something more to
your post, David, yet those are its last two lines. What's up with that?
Patty Winter
In article <4d5e0933$1...@news.x-privat.org>,
Here's the link that the article probably had:
http://www.pbs.org/wgbh/nova/tech/smartest-machine-on-earth.html
Patty
Padmar Mushkin
>That's because, Breeden writes, "Watson had an unfair advantage: It
>was being fed the questions electronically. I wanted to see Watson
>hear the questions using speech recognition and process them the way
>humans do.
It didn't even have to do that. It would have been easy enough to set
up a camera and have Watson read the clues.
Goro
>... not
> much different than what Google and Bing do every day right now."
I dislike this statement very much. Pagerank is very different from
natural language analysis and machine learning.
-goro-
tdciago
> It didn't even have to do that. It would have been easy enough to set
> up a camera and have Watson read the clues.
Watson can't see or hear, which is why there were no video clues (such
as images of paintings) or clues that relied on audio (such as playing
certain tunes).
Mason Barge
Yeah, and you still have your heels dug into an incorrect position. Try
to understand the point he's making.
Mason Barge
><spoilers below>
>
>*
>
>*
>
>*
>
>*
>
>*
>
>*
>
>*
>
>That's because, Breeden writes, "Watson had an unfair advantage: It
>was being fed the questions electronically.
Anyone who works with computers and watches Jeopardy is going to realize
that I/O factors are going to make the comparison of "computing" power
invalid.
But, IMHO, the contest didn't even reach the substantive "I" part of the
I/O problem. Watson would have still won, IMHO, even if it had received
the questions on a more equal footing, because of its ability to buzz in
.0001 second after the "go" light turned on.
Dano
"Mason Barge" <mason...@gmail.com> wrote in message
news:r1dtl65uu6rvq3l62...@4ax.com...
I have the solution. Watson should need to "boot up" in order to answer
each question. Since it takes a human to turn it on, then it would be a
"fair fight".
Mason Barge
I'd hate to see what happens when they don't let the contestants use alarm
clocks.
Barry Margolin
> On Thu, 17 Feb 2011 23:28:42 -0500, David wrote:
>
> > Writing today, Feb. 17, 2011 on GCN's website, John Breeden II said,
> > "You would think that a techie like me would be overjoyed at seeing a
> > glimpse into the rise [of] what show loser Ken Jennings
>
> It says a lot about the state of our economy that someone who won
> $300,000 is considered to be a "loser". :)
It's not an judgement call about Ken, just a descriptive fact. He
didn't win the game; he lost, so he's a loser.
--
Barry Margolin, bar...@alum.mit.edu
Arlington, MA
*** PLEASE don't copy me on replies, I'll read them in the group ***
Barry Margolin
"Dano" <janea...@yahoo.com> wrote:
I have a different solution. Don't play a game with a race to answer.
Allow every contestant to answer within a short time, and just compare
them on the number of right answers.
We can still use Jeopardy's clues, but the Jeopardy game format is
inappropriate if you want a contest of minds, not fingers.
Alan Curry
Barry Margolin <bar...@alum.mit.edu> wrote:
>
>I have a different solution. Don't play a game with a race to answer.
>Allow every contestant to answer within a short time, and just compare
>them on the number of right answers.
That was the bonus round of Win Ben Stein's Money.
--
Alan Curry
Stan Brown
> I have a different solution. Don't play a game with a race to answer.
> Allow every contestant to answer within a short time, and just compare
> them on the number of right answers.
>
> We can still use Jeopardy's clues, but the Jeopardy game format is
> inappropriate if you want a contest of minds, not fingers.
But Jeopardy has been a contest of fingers *and* minds for years. I
don't remember how it was in the Fleming days, but AFAIK under Trebek
the buzzers have been locked out while he reads for many years. The
game wasn't about whether Watson was "smarter" than humans, but about
whether it was better at Jeopardy.
I am still troubled by the input issue by which Watson gets the
question in a text file instead of human speech to be recognized. I
hope in a few years there is another match where speech recognition
is part of the setup.
But the buzzer business seems to me like a red herring. Even for
humans a big part of success is proficiency with the buzzer. And
they didn't let Watson buzz in electronically; it had to press a
physical buzzer, and was sometimes beat to it by a human.
--
Stan Brown, Oak Road Systems, Tompkins County, New York, USA
http://OakRoadSystems.com
"Children -- so adorable. In a way they're like people."
-- Veronica, on /Better Off Ted/
Patty Winter
In article <MPG.27c98be77...@news.individual.net>,
Stan Brown <the_sta...@fastmail.fm> wrote:
>On Fri, 18 Feb 2011 21:51:20 -0500, Barry Margolin wrote:
>
>> I have a different solution. Don't play a game with a race to answer.
>> Allow every contestant to answer within a short time, and just compare
>> them on the number of right answers.
>>
>> We can still use Jeopardy's clues, but the Jeopardy game format is
>> inappropriate if you want a contest of minds, not fingers.
>
>But Jeopardy has been a contest of fingers *and* minds for years. I
>don't remember how it was in the Fleming days, but AFAIK under Trebek
>the buzzers have been locked out while he reads for many years. The
>game wasn't about whether Watson was "smarter" than humans, but about
>whether it was better at Jeopardy.
Then I, too, would like to see Watson engage in a natural-language
contest in which all contestants are allowed to offer answers.
Although I doubt that the Jeopardy producers would have allowed this
one to proceed unless they believed that the buzzer response time
issue had been addressed. (Perhaps by the fact that the humans could
anticipate when Alex would finish speaking. I realize, however, that
there is still controversy about whether that actually leveled the
playing field.)
Patty
Barry Margolin
> On Fri, 18 Feb 2011 21:51:20 -0500, Barry Margolin wrote:
>
> > I have a different solution. Don't play a game with a race to answer.
> > Allow every contestant to answer within a short time, and just compare
> > them on the number of right answers.
> >
> > We can still use Jeopardy's clues, but the Jeopardy game format is
> > inappropriate if you want a contest of minds, not fingers.
>
> But Jeopardy has been a contest of fingers *and* minds for years.
So? Why does it have to be Jeopardy?
The only reason is because it's a popular trivia game show. But if we
want to have a decent human-computer challenge, we don't need to use
that exact format.
Stan Brown
Like you, I believe that the producers were satisfied with the
handling of the buzzer. If not, I can't imagine they would have
agreed to the show.
We should also remember that Jennings and Rutter had some number of
practice sessions with Watson, so presumably they had time to get
used to how Watson handles the buzzer. If I recall correctly from
Nova, they used to eat it handily.
Stan Brown
>
> In article <MPG.27c98be77...@news.individual.net>,
> Stan Brown <the_sta...@fastmail.fm> wrote:
>
> > On Fri, 18 Feb 2011 21:51:20 -0500, Barry Margolin wrote:
> >
> > > I have a different solution. Don't play a game with a race to answer.
> > > Allow every contestant to answer within a short time, and just compare
> > > them on the number of right answers.
> > >
> > > We can still use Jeopardy's clues, but the Jeopardy game format is
> > > inappropriate if you want a contest of minds, not fingers.
> >
> > But Jeopardy has been a contest of fingers *and* minds for years.
>
> So? Why does it have to be Jeopardy?
>
> The only reason is because it's a popular trivia game show. But if we
> want to have a decent human-computer challenge, we don't need to use
> that exact format.
"Have to use"? No, of course not. But it's a part of the zeitgeist,
even for people who don't watch it.
This seems to me to be a good analogy to the chess-playing programs
of a few decades ago. Everyone, even the majority who wouldn't know
a bishop from a pawn, thinks of chess as a game that displays
intelligence. So it was natural to use chess as a measure of
artificial intelligence.
People also tend to speak of Jeopardy as a game for smart people --
even though most of us know that raw intelligence is a small part of
the recipe for success. So it seemed like a natural for this year's
test of artificial intelligence.
I hope we can all agree that, whatever qualms we may have about the
details, it was darned good television.
tdciago
> So? Why does it have to be Jeopardy?
It doesn't have to be "Jeopardy!", but this time around it was.
Removing the buzzer from the equation would've meant it was no longer
"Jeopardy!" The idea behind this particular challenge arose during
Ken's original run, and IBM was able to see it through to a
spectacular conclusion.
They can certainly conceive of new challenges for Watson now.
Mason Barge
I'd say that anticipating an end to the question is a legitimate cognitive
function that Watson should fairly be expected to imitate.
Mason Barge
<the_sta...@fastmail.fm> wrote:
Nah. The lack of a level playing field ruined it for me, personally.
Patty Winter
In article <len0m6haf06vkj891...@4ax.com>,
Mason Barge <mason...@gmail.com> wrote:
>On 19 Feb 2011 20:13:07 GMT, Patty Winter <pat...@wintertime.com> wrote:
bunch of unneeded quotage removed
>>Although I doubt that the Jeopardy producers would have allowed this
>>one to proceed unless they believed that the buzzer response time
>>issue had been addressed. (Perhaps by the fact that the humans could
>>anticipate when Alex would finish speaking. I realize, however, that
>>there is still controversy about whether that actually leveled the
>>playing field.)
>
>I'd say that anticipating an end to the question is a legitimate cognitive
>function that Watson should fairly be expected to imitate.
Huh? Are you saying that Watson should have been programmed to
calculate how long Alex Trebek would take to read the clue just
sent to it?
Patty
Joshua Kreitzer
>
> http://www.tvweek.com/blogs/tvbizwire/2011/02/why-former-jeopardy-champ-ken.php
>
> According to a story in The New York Times, IBM researchers
> acknowledged that Watson benefited from the "buzzer factor," because
> its human competitors had to anticipate the light that signals it's
> possible to "buzz in" to provide a response, while the computer uses a
> weighted scheme that allows it to buzz in as quickly as 10
> milliseconds. That made it hard for the human competitors to hit their
> buzzers in time.
I think the buzzer factor is kind of a red herring. It's not like the
primary challenge for IBM was to develop a computer system that could
beat humans to the buzzer -- that was probably the easy part. The
challenge was to develop a computer system that could answer the
Jeopardy clues. Few people would have been impressed if IBM developed
a computer that buzzed in first every time but got most of the
responses wrong.
--
Joshua Kreitzer
grom...@hotmail.com
Stan Brown
Thanks. You said more briefly and better what I was trying to say at
greater length.
Mason Barge
If that's what the contestants do, then yes. (And I believe they do
sometimes.)
Mason Barge
True enough. But still, Watson was playing on an uneven field. I think
you could also say that part of the task was to build a computer that
could answer the questions better than the best humans, and that was not
fairly tested.
Barry Margolin
Mason Barge <mason...@gmail.com> wrote:
Me, too. It didn't take long to see that it would be an incredibly
lop-sided competition. A game show with little uncertainty about the
winner isn't much fun to watch.
Obveeus
"Barry Margolin" <bar...@alum.mit.edu> wrote:
> Me, too. It didn't take long to see that it would be an incredibly
> lop-sided competition. A game show with little uncertainty about the
> winner isn't much fun to watch.
You watch trivia gameshows to see who will win? I don't really care who
wins. If I watch a show like Jeopardy, Cash Cab, or even Wheel of Fortune,
it is so that I can play along and try to answer the questions myself.
Barry Margolin
Mason Barge <mason...@gmail.com> wrote:
> On 20 Feb 2011 00:49:32 GMT, Patty Winter <pat...@wintertime.com> wrote:
>
> >
> >In article <len0m6haf06vkj891...@4ax.com>,
> >Mason Barge <mason...@gmail.com> wrote:
> >>On 19 Feb 2011 20:13:07 GMT, Patty Winter <pat...@wintertime.com> wrote:
> >
> >
> >bunch of unneeded quotage removed
> >
> >
> >>>Although I doubt that the Jeopardy producers would have allowed this
> >>>one to proceed unless they believed that the buzzer response time
> >>>issue had been addressed. (Perhaps by the fact that the humans could
> >>>anticipate when Alex would finish speaking. I realize, however, that
> >>>there is still controversy about whether that actually leveled the
> >>>playing field.)
> >>
> >>I'd say that anticipating an end to the question is a legitimate cognitive
> >>function that Watson should fairly be expected to imitate.
> >
> >
> >Huh? Are you saying that Watson should have been programmed to
> >calculate how long Alex Trebek would take to read the clue just
> >sent to it?
>
> If that's what the contestants do, then yes. (And I believe they do
> sometimes.)
Maybe sometimes, but I think most of the time they're listening to him.
They've already read the entire clue and figured out the answer. They
can hear when he's saying the last word, and know that the buzzer will
be enabled as soon as he finishes the word.
Only a deaf contest (like Watson) would need to guess Alex's entire
clue-reading time.
It probably wouldn't be too hard to add that ability to Watson. It
wouldn't need complete real-time speech recognition software. All it
has to do is listen for the last couple of words in the clue.
suzeeq
Same here.
Barry Margolin
"Obveeus" <Obv...@aol.com> wrote:
That's a normal way to watch game shows, but not a special event like
this one. This one was supposed to be interesting because of the
man-vs-machine element.
Except for the categories that Watson didn't "get", it became pretty
uninteresting after a while.
I was impressed at how good Watson was, since the version on the Nova
documentary was so far from being ready to compete. But it was so good
that it was no longer riveting television, since there was no tension or
drama. They might as well have put Watson on "Deal or No Deal".
Remysun
> That's a normal way to watch game shows, but not a special event like
> this one. This one was supposed to be interesting because of the
> man-vs-machine element.
>
> Except for the categories that Watson didn't "get", it became pretty
> uninteresting after a while.
>
> I was impressed at how good Watson was, since the version on the Nova
> documentary was so far from being ready to compete. But it was so good
> that it was no longer riveting television, since there was no tension or
> drama. They might as well have put Watson on "Deal or No Deal".
I agree. Watson was more exciting when it was "within the cloud".
tdciago
> It didn't take long to see that it would be an incredibly
> lop-sided competition. A game show with little uncertainty about the
> winner isn't much fun to watch.
That's how I felt during Ken's run, and it's one reason I dislike the
elimination of the 5-game limit. Watson, over time, learned strategies
like hunting for Daily Doubles, and had its database expanded. A human
contestant also benefits from increased confidence and familiarity
with the set and equipment and being on TV. If a contestant like Ken
is permitted to continue to compete as his game grows stronger, it
wouldn't be fair to stop Watson from competing because it got so good
at the game. Once Watson reached the point of regularly beating
humans, including former contestants who had not necessarily performed
at Ken's or Brad's level, would it have been fair to demand that
Watson face those two before it got any better? I don't think so.
The point of this challenge was not to protect the fragile human
ego.
No Jeopardy! game is ever played by three contestants with totally
equal skills. If this 2-game challenge proved anything, it's that
Watson is *not* unbeatable. We saw in the first game that Watson is
vulnerable in Final Jeopardy as long as a lock game can be prevented.
In the second game, Ken managed to do exactly that, but the result in
that case wasn't combined with a FJ miss by Watson. When Ken
originally played, it took 75 games before that particular combination
of events happened, but ultimately it *did* happen. I suspect it
would have taken far less time for Ken and Brad to hit that combo with
Watson.
Remember that the game is not always won by the contestant who excels
at responding correctly in the first two rounds. Daily Doubles, key
wagering decisions, and that pesky Final Jeopardy clue can all bring
down a player who has dominated throughout the game. (Hence the title
of the show.)
My second game is a perfect example of this. In no way am I claiming
to be *anywhere* near the level of the best J! players, but this is
simply an illustration of what can happen in any game. These are the
Coryat scores, which are the scores based solely on the face value of
clues in the first two rounds, without regard to extra money won or
lost in Daily Doubles.
Ronnie (that's me) - $19,600 (28 right, including 2 DD; 2 wrong)
Allison - $12,000 (17 right, including 1 DD; 4 wrong)
Ben - $4,600 (8 right; 4 wrong)
Ben won that game, and went on to become a 5-time champ. Allison
wagered $6,000 on a Daily Double clue normally worth $1,200, and got
it right. The scores going into FJ were:
Ronnie - $19,900
Allison - $16,800
Ben - $4,600
Ben was the only player to get FJ right. Had Allison not earned that
extra $4,800 on the DD, I wouldn't have had to wager so much to cover
her in FJ, and it would have been impossible for Ben to win. So we
see how that combo of DD, wagering, and FJ all contributed to my loss.
These are the Coryat scores in Watson's second game:
Watson - $26,800 - 28 R (including 1 DD), 3 W (including 1 DD)
Ken - $15,200 - 17 R (including 1 DD), 1W
Brad - $5,600 - 12 R, 2 W
Scores entering FJ were:
Watson - $23,440
Ken - $18,200
Brad - $5,600
Ken was well within range of winning this particular game had Watson
missed FJ, as it did in the first game.
A brief 2-game challenge does not demonstrate that Watson is too good
to lose, anymore than Ben's overall performance in the game above
indicated his true ability.
Mason Barge
Yeah it's part of the fun. I'm with you halfway :) But if there's
someone I like or someone that rubs me the wrong way, I definitely have
fun pulling for someone.
And actually, I don't really want to watch any contest where both factors
aren't present. I like the Amazing Race because I love the travelogue
aspect, the competitions are often fun to watch, *and* it's interesting
when personalities of the competitors affect how well they do. Like if
they have to scuba dive (last night) but some of the people are
claustraphobic or afraid of water.
And I definitely pull for some teams to win.
Mason Barge
>On Feb 20, 8:59 pm, Barry Margolin <bar...@alum.mit.edu> wrote:
>> It didn't take long to see that it would be an incredibly
>> lop-sided competition. A game show with little uncertainty about the
>> winner isn't much fun to watch.
>
>That's how I felt during Ken's run, and it's one reason I dislike the
>elimination of the 5-game limit. Watson, over time, learned strategies
>like hunting for Daily Doubles, and had its database expanded.
In fairness to Ken Jennings, I don't think he ever engaged in daily double
hunting.
It is interesting to watch which contestants do and don't. There is
certainly (at least for some people) a feeling that breaking a category
other than with the lowest value question is rude and unsportsmanlike.
I usually pull against people who do it, and I bet a lot of other people
do, too.
tdciago
> In fairness to Ken Jennings, I don't think he ever engaged in daily double
> hunting.
Right, I didn't mean to imply that Ken had done that particular
thing. I was just pointing out an example of a strategy that Watson
learned to employ as it became more experienced at playing the game.
Ken developed and sharpened his own strategies over time. Obviously
he decided to switch things up and hit the high-value clues right off
the bat in the Double Jeopardy round of the first game against Watson,
but that was not his typical style of play.
> It is interesting to watch which contestants do and don't.
Most of the time, the people who decide to pull this in their first
game fail spectacularly. I recall one guy who ended up deep in the
red and wasn't even around for Final Jeopardy. It's very
disconcerting, which is part of the reason people attempt it, but they
end up being just as screwed up as their opponents. Watson is not
subject to the feeling of disorientation that can result from being
thrown into the hardest clues without gradually leading up to them, so
it's a sensible strategy for it to use.
Ken Wesson
> In article <4d5e0933$1...@news.x-privat.org>,
> Ken Wesson <kwe...@gmail.com> wrote:
>
>> On Thu, 17 Feb 2011 23:28:42 -0500, David wrote:
>>
>> > Writing today, Feb. 17, 2011 on GCN's website, John Breeden II said,
>> > "You would think that a techie like me would be overjoyed at seeing a
>> > glimpse into the rise [of] what show loser Ken Jennings
>>
>> It says a lot about the state of our economy that someone who won
>> $300,000 is considered to be a "loser". :)
>
> It's not an judgement call about Ken, just a descriptive fact. He
> didn't win the game; he lost
He came in second. Brad "lost" (and even he came away with $200,000).
Barry Margolin
<30f44ebe-3c46-411b...@r19g2000prm.googlegroups.com>,
tdciago <tdc...@aol.com> wrote:
> On Feb 21, 1:31 pm, Mason Barge <masonba...@gmail.com> wrote:
> > In fairness to Ken Jennings, I don't think he ever engaged in daily double
> > hunting.
>
> Right, I didn't mean to imply that Ken had done that particular
> thing. I was just pointing out an example of a strategy that Watson
> learned to employ as it became more experienced at playing the game.
Watson didn't "learn" that. The human programmers noticed that Daily
Doubles are usually in the bottom two rows, so they deliberately
programmed this strategy into it.
Michael Black
> In article
> <30f44ebe-3c46-411b...@r19g2000prm.googlegroups.com>,
> tdciago <tdc...@aol.com> wrote:
>
>> On Feb 21, 1:31 pm, Mason Barge <masonba...@gmail.com> wrote:
>>> In fairness to Ken Jennings, I don't think he ever engaged in daily double
>>> hunting.
>>
>> Right, I didn't mean to imply that Ken had done that particular
>> thing. I was just pointing out an example of a strategy that Watson
>> learned to employ as it became more experienced at playing the game.
>
> Watson didn't "learn" that. The human programmers noticed that Daily
> Doubles are usually in the bottom two rows, so they deliberately
> programmed this strategy into it.
>
Which is a difference between people and computers. We adjust to
things (well some of the time) while a computer keeps on going until
the program changed.
Surely for artificial intelligence to happen, it requires an ability
to self-program, so it can add that bit of code to watch for something
specific. Of course, then it requires an ability to actually learn from
the experience in order to recode itself (and by "learning" I mean
reacting to something and changing, not "stuffing oneself with more
trivia".
Michael
chicagofan
> On Tue, 22 Feb 2011, Barry Margolin wrote:
>
>> tdciago<tdc...@aol.com> wrote:
>>
>>> On Feb 21, 1:31�pm, Mason Barge<masonba...@gmail.com> wrote:
>>>
>>>> In fairness to Ken Jennings, I don't think he ever engaged in daily double
>>>> hunting.
>>>>
>>> Right, I didn't mean to imply that Ken had done that particular
>>> thing. I was just pointing out an example of a strategy that Watson
>>> learned to employ as it became more experienced at playing the game.
>>>
>> Watson didn't "learn" that. The human programmers noticed that Daily
>> Doubles are usually in the bottom two rows, so they deliberately
>> programmed this strategy into it
> things (well some of the time) while a computer keeps on going until
> the program changed.
>
> Surely for artificial intelligence to happen, it requires an ability
> to self-program, so it can add that bit of code to watch for something
> specific. Of course, then it requires an ability to actually learn from
> the experience in order to recode itself (and by "learning" I mean
> reacting to something and changing, not "stuffing oneself with more
> trivia".
>
According to the Nova program I just watched recently on PBS about
"Watson", it did indeed learn some things from experience. Can't tell
you how/which things ... there was just too much to absorb and remember.
:) Very interesting show about the programmers and it's
creation/development.
bj
Barry Margolin
chicagofan <chica...@privacy.net> wrote:
It was supposed to learn about categories from its mistakes.
chicagofan
> chicagofan<chica...@privacy.net> wrote:
>
>> According to the Nova program I just watched recently on PBS about
>> "Watson", it did indeed learn some things from experience. Can't tell
>> you how/which things ... there was just too much to absorb and remember.
>> :) Very interesting show about the programmers and it's
>> creation/development.
>> bj
>>
> It was supposed to learn about categories from its mistakes.
>
Thanks! Sorry, if that had already been discussed. I haven't been able
to keep up with all the threads/messages lately. :)
bj