[Dixonary] OT: Jeopardy

[Judy Madnick]

From: "Dodi Schultz" <DodiSchultz (AT) nasw (DOT) org>

<< As a faithful "Jeopardy" follower, I've seen Jennings and Rutter
<< perform
<< before, and I'm totally sure they've both known most of the
<< answers
<< (geez, *I've* known most of them). Watson's way ahead only
<< because the
<< electronic "ring-in" can move way faster than the human
<< thumb. It's not
<< an intelligence competition; it's a reaction contest, and it's
<< measured
<< in nanoseconds. Maybe picoseconds.

I agree. And I had a feeling that Watson wouldn't get the Final Jeopardy question based on the nature of the question -- and he was way off! Even *I* guessed right!! I found that I could kind of sense the types of questions that Watson would miss, did you? The Final Jeopardy question was a perfect example of one that Watson would be unable to process correctly.

Judy

[Steve Graham]

IBM obviously put a lot of work into the Watson project (as opposed to the
ginormous screwup they perpetrated on the 1996 Olympics in Atlanta
processing and reporting results.)


Here's an interesting discussion about how Watson blew the final question
Tuesday
http://asmarterplanet.com/blog/2011/02/watson-on-jeopardy-day-two-the-confus
ion-over-an-airport-clue.html

[France International]

I agree that the game was "rigged" due to Watson's instantaneous reaction
time (I've heard that some contestants attempt to anticipate the
"ring-in-allowed" light, but that's a dangerous strategy, because if you
miss, you are penalized by not being able to ring in again for a short
period of time). That notwithstanding, I was impressed with Watson's ability
to come up with the correct answer more that 90% of the time. That's quite a
programming feat, if you think of all the variables that have to be
considered and associations that have to be made. On the other hand, Watson
gave a few surprisingly stupid answers, particularly on that final jeopardy
question - the clue asked for US city after all. It reminded me of the
early chess programs that could calculate 15 or 20 moves ahead, but for some
reason would make a dumb move and lose the game.

--Mike

----- Original Message -----
From: "Judy Madnick" <jmadnick (AT) gmail (DOT) com>
To: <dixonary (AT) googlegroups (DOT) com>
Sent: Wednesday, February 16, 2011 10:13 AM
Subject: [Dixonary] OT: Jeopardy


> From: "Dodi Schultz" <DodiSchultz (AT) nasw (DOT) org>
>
> << As a faithful "Jeopardy" follower, I've seen Jennings
> and Rutter
> << perform
> << before, and I'm totally sure they've both known most of
> the
> << answers
> << (geez, *I've* known most of them). Watson's way ahead
> only
> << because the
> << electronic "ring-in" can move way faster than the human
> << thumb. It's not
> << an intelligence competition; it's a reaction contest,
> and it's
> << measured
> << in nanoseconds. Maybe picoseconds.
>
> I agree. And I had a feeling that Watson wouldn't get the Final Jeopardy
> question based on the nature of the question -- and he was way off! Even
> *I* guessed right!! I found that I could kind of sense the types of
> questions that Watson would miss, did you? The Final Jeopardy question was
> a perfect example of one that Watson would be unable to process correctly.
>
> Judy

[France International]

Interesting discussion, especially about potential medical and legal uses.
Maybe in 20 years we'll see a version of Watson on the Supreme Court (well
not really, but I wouldn't be at all surprised if something like it was used
in an advisory capacity).

--Mike

----- Original Message -----
From: "Steve Graham" <sdgraham (AT) duckswild (DOT) com>
To: <dixonary (AT) googlegroups (DOT) com>
Sent: Wednesday, February 16, 2011 10:35 AM
Subject: RE: [Dixonary] OT: Jeopardy


> IBM obviously put a lot of work into the Watson project (as opposed to the
> ginormous screwup they perpetrated on the 1996 Olympics in Atlanta
> processing and reporting results.)
>
>
> Here's an interesting discussion about how Watson blew the final question
> Tuesday
> http://asmarterplanet.com/blog/2011/02/watson-on-jeopardy-day-two-the-confus
> ion-over-an-airport-clue.html
>

[Dodi Schultz]

Steve Graham wrote:

> Here's an interesting discussion about how Watson blew the final question
> Tuesday
> http://asmarterplanet.com/blog/2011/02/watson-on-jeopardy-day-two-the-confus
> ion-over-an-airport-clue.html
>

Interesting; thanks for the link, Steve!

[Judy Madnick]

From: "France International" <stamps (AT) salsgiver (DOT) com>

<< I agree that the game was "rigged" due to Watson's
<< instantaneous reaction
<< time

Do you think it was actually "rigged" -- or do you believe it was just the advantage of being a computer?

It was amazing to observe Watson's ability to answer correctly so often. And when "he" was wrong, I think you could often tell by the nature of the question why "he" couldn't arrive at the correct answer.

All in all, it was fun.

Judy

[—Keith Hale—]

Yes, thanks for the link! (I have a TV, but can only use it as a DVD or
computer monitor, so i've not seen any of this.) Deep Blue, eat your CPU
out!

Here's an interesting discussion about how Watson blew the final question
>> Tuesday
>>
>> http://asmarterplanet.com/blog/2011/02/watson-on-jeopardy-day-two-the-confus
>> ion-over-an-airport-clue.html
>>
>
> Interesting; thanks for the link, Steve!
>

[Dave Cunningham]

The timing on the buzzer is indeed key (having been tested twice by
now by me) and even a 50 ms edge on the computer "reflexes" is enough
to highly skew the results. They make a point that the computer uses a
relay to press the buzzer - but that is not the real point. The real
lesson is that the computer is still fully capable of extremely wrong
answers ("Toronto"?).

The program is primarily one which comes closer to how people think
than the old chess programs - it looks for patterns of related words,
and chooses from those words or phrases "most closely related" to all
the words it identifies as "key words" in the answer (and using the
"topic" as an additional common "key word" and noting likely reasons
for the wording of the topic -- quote marks mean it has those letters
in the answer, etc.)

This is better than the old "flow chart" expert systems, but has a
very long way to go overall.

Dave (who took the online test again)

On Feb 16, 10:13*am, "Judy Madnick" <jmadn... (AT) gmail (DOT) com> wrote:
> * * * * * * * * *From: "Dodi Schultz" <DodiSchu... (AT) nasw (DOT) org>
>
> * * * * * * * * << As a faithful "Jeopardy" follower, I've seen Jennings and Rutter
> * * * * * * * * << perform
> * * * * * * * * << before, and I'm totally sure they've both known most of the
> * * * * * * * * << answers
> * * * * * * * * << (geez, *I've* known most of them). Watson's way ahead only
> * * * * * * * * << because the
> * * * * * * * * << electronic "ring-in" can move way faster than the human
> * * * * * * * * << thumb. It's not
> * * * * * * * * << an intelligence competition; it's a reaction contest, and it's
> * * * * * * * * << measured
> * * * * * * * * << in nanoseconds. Maybe picoseconds.
>
> I agree. And I had a feeling that Watson wouldn't get the Final Jeopardy question based on the nature of the question -- and he was way off! Even *I* guessed right!! I found that I could kind of sense the types of questions that Watson would miss, did you? The Final Jeopardy question was a perfect example of one that Watson would be unable to process correctly.
>
> Judy

[Dave Cunningham]

Counting the times Watson did not even ring in, or showed the "wrong
answer" as one it was considering, its actual accuracy was more like
75% - which is just about what most good Jeopardy players average.
And it does not try to parse sentences - it looks at key words and
words which are then shared in connection to those key words.

On Feb 16, 10:52*am, "France International" <sta... (AT) salsgiver (DOT) com>
wrote:
> I agree that the game was "rigged" due to Watson's instantaneous reaction
> time (I've heard that some contestants attempt to anticipate the
> "ring-in-allowed" light, but that's a dangerous strategy, because if you
> miss, you are penalized by not being able to ring in again for a short
> period of time). That notwithstanding, I was impressed with Watson's ability
> to come up with the correct answer more that 90% of the time. That's quite a
> programming feat, if you think of all the variables that have to be
> considered and associations that have to be made. On the other hand, Watson
> gave a few surprisingly stupid answers, particularly on that final jeopardy
> question - the clue asked for *US city after all. It reminded me of the
> early chess programs that could calculate 15 or 20 moves ahead, but for some
> reason would make a dumb move and lose the game.
>
> --Mike
>
>
>
> ----- Original Message -----
> From: "Judy Madnick" <jmadn... (AT) gmail (DOT) com>
> To: <dixonary (AT) googlegroups (DOT) com>
> Sent: Wednesday, February 16, 2011 10:13 AM
> Subject: [Dixonary] OT: Jeopardy
>
> > * * * * * * * * From: "Dodi Schultz" <DodiSchu... (AT) nasw (DOT) org>
>
> > * * * * * * * *<< As a faithful "Jeopardy" follower, I've seen Jennings
> > and Rutter
> > * * * * * * * *<< perform
> > * * * * * * * *<< before, and I'm totally sure they've both known most of
> > the
> > * * * * * * * *<< answers
> > * * * * * * * *<< (geez, *I've* known most of them). Watson's way ahead
> > only
> > * * * * * * * *<< because the
> > * * * * * * * *<< electronic "ring-in" can move way faster than the human
> > * * * * * * * *<< thumb. It's not
> > * * * * * * * *<< an intelligence competition; it's a reaction contest,
> > and it's
> > * * * * * * * *<< measured
> > * * * * * * * *<< in nanoseconds. Maybe picoseconds.
>
> > I agree. And I had a feeling that Watson wouldn't get the Final Jeopardy
> > question based on the nature of the question -- and he was way off! Even
> > *I* guessed right!! I found that I could kind of sense the types of
> > questions that Watson would miss, did you? The Final Jeopardy question was
> > a perfect example of one that Watson would be unable to process correctly.
>
> > Judy- Hide quoted text -
>
> - Show quoted text -

[Dodi Schultz]

Judy Madnick wrote:
> From: "France International" <stamps (AT) salsgiver (DOT) com>
>
> << I agree that the game was "rigged" due to Watson's
> << instantaneous reaction
> << time
>
> Do you think it was actually "rigged" -- or do you believe it was just the advantage of being a computer?
>

"Rigged" in the sense that it's by no means the touted test of embryonic
AI we'd been led to expect. If "Watson" had been slowed to human
thumb-activation time, it might have been.

I found it disappointing.

[Dodi Schultz]

---Keith Hale--- wrote:

> Yes, thanks for the link! (I have a TV, but can only use it as a DVD
> or computer monitor, so i've not seen any of this.)

You have no cable or other service in your area?

Anyway: Aren't most TV shows now available later on the Web? Not that I
choose to watch any that way, but . . .

[—Keith Hale—]

As a starving art student i don't do cable, and have never been able to make
any digital conversion antenna rig work. I have netflix and internet, but
the latter is so bad i gave up on most streaming possibilities.

The "Toronto" answer isn't the right one, of course, but there are US cities
named thus, and i wouldn't have guessed Chicago in my top 10 answers. Also
- it bet intelligently, so it still won! Was it playing for a charity,
maybe? Where does the money it wins go?


On 16 February 2011 13:43, Dodi Schultz <DodiSchultz (AT) nasw (DOT) org> wrote:

> —Keith Hale— wrote:
>
> Yes, thanks for the link! (I have a TV, but can only use it as a DVD or
> computer monitor, so i've not seen any of this.)
>
>
> You have no cable or other service in your area?
>
> Anyway: Aren't most TV shows now available later on the Web? Not that I
> choose to watch any that way, but . . .
>