Watson and Jeopardy

[Lennart Sorensen]

On Wed, Feb 16, 2011 at 04:35:26PM -0500, Yanni Chiu wrote:
> On 16/02/11 11:07 AM, John Myshrall wrote:
>>
>> I wonder if Watson was joking or being merciful if that is even remotely
>> possible or that someone on IBM intervened.
>
> I had the same feeling - that some person put the brakes on. With so  
> many right answers prior, it seems impossible that its best guess would  
> be Toronto. Any U.S. city with two airports would have been a better 
> guess.
>
> There weren't many opportunities to observe the response when the  
> machine is actually stumped. Maybe this is the fail safe response - try  
> and be funny.
>
>> Toronto was so far out in left field it fails on the first litmus test.
>> The answer was followed by many question marks and it only wager
>> <$1000.00 yet it had a $25000.00 lead.
>
> I thought the final wager was way too low. The previous "double  
> jeopardy" wagers seemed too low, as well. I don't think any human player  
> would have bet so low, knowing how the game has been going, and their  
> own abilities.

Remember that for final jeopardy and daily doubles, you have to bet
before you know the clue.  That's risky.  With normal clues you only
have to buzz in and take a chance if you are fairly sure of your answer.

The final jeopardy bet was actually very much in line with how humans
often bet when they have a huge lead.

> I didn't see the first show, so I can't figure out how the humans could  
> have been so close in score from the previous day. It seemed to me that  
> some fine tuning had gone on between the first day and the second day,  
> to maximize the response time. It seemed to me that the computer had  
> super-human ability to hit the buzzer as soon as it had it's response 
> ready.

-- 
Len Sorensen
--
The Toronto Linux Users Group.      Meetings: http://gtalug.org/
TLUG requests: Linux topics, No HTML, wrap text below 80 columns
How to UNSUBSCRIBE: http://gtalug.org/wiki/Mailing_lists