Men v. Machine(s): The Rest of the Story?

By James Fallows

by Edward Goldstick

First, this is a "prologue" that I wrote on Monday but that did not get posted (I didn't understand the process yet):

I didn't get home fast enough to do more than enter my "guess" before the fireworks start ... so here goes: with one caveat, I expect that Watson will win decisively. The caveat is that there will be no explicit delay built into the "button" for the machine. More on this in a "real" post in a moment... but I want to save this for posterity. @19:29 Eastern on Monday, February 14, 2010

Now that the results are known, it seems superfluous to recount what has transpired; instead, a few personal conjectures that reflect my three biggest surprises after watching (and after a bit of further reading):

1. There apparently was, in fact, an electronic servomechanism that mechanically pushed the button for Watson, but did it contain a delay equivalent to a human reflex time?

Is this a new definition of "being human?" Though I have watched Jeopardy! since I was a wee lad, this puts a new light on the notion that "finding the right question" is often more important than elucidating an answer.

2. The arbitration of results from multiple, while apparently distinct, algorithms that "add up" to a confidence factor for a "correct" answer (or "question," in Jeopardy! parlance) ... but then again, I do not remember seeing any mention of "translation" from one human tongue to another as a "natural" application of Watson's computing power.

3. Ken and Brad really didn't stand a chance ... and yes, I was rooting for them!

The Big Question, of course, is the one posed by Brian Christian in his fascinating article in the current edition of The Atlantic [(though I don't get paid for this... subscribe!]: Does this constitute a valid Turing Test? I say "yes and no" on the following terms:

- Yes if we consider the few "dumb" answers (Toronto in the U.S.?) though players on Jeopardy! sometimes do put up wisecrack answers in Final Jeopardy when they are befuddled. And Watson did, after all, bet very little!

- No quite simply because Watson did not participate in the banter or answer any questions during the interludes. Or maybe it is just shy?

As in the previous post on the "IBM/Jeopardy! Challenge", here are "running notes" that I took while watching Days Two and Three...

Day Two: 

IBM publicity:
... The Apollo Project:  "IBM played a big part..." - well, so did public financing, no?
"Open Question Answering"
"When we pursue natural language, we pursue the heart of human intelligence"...
... Some basic thoughts:
Is the ambiguity of the category as complicated/important as anything/everything else?
Just how much does Watson continue to process before the first question and between questions?

And now, DOUBLE JEOPARDY...

3 / 23 / 2 ... 2

... and Sheesh, Watson is so far ahead in $$ that I didn't note it down...

Some random thoughts and observations before Final Jeopardy:

... is there a "warming up" time for the human brain?
... MY WIFE GOT "MODERN ART" when all three guessed something and got it wrong!
... "Information overload is the problem of our day..." (don't be such a wet towel...)
... no mention of translation in the IBM ads (so far)
... How can the computer fail at Final Jeopardy if it now has 30 seconds to guess.

And now, the first FINAL JEOPARDY...

... US City w/ airports named for WWII hero and WWII battle = O'Hare and Midway = CHICAGO
... Category is "U.S. Cities"... and Watson selects "TORONTO????"
... but when in the lead and in doubt, don't tempt it!

Day Three:

JEOPARDY:  10 / 11 / 8 ... 2

... There sure seemed to be some simple questions that he... eh, it... came up blank...
... Did they "turn down" Watson to make it more competitive?

DOUBLE JEOPARDY:  7 / 18  / 4 ... 0

... Apparently not.

FINAL JEOPARDY: 

... Ken has about $18000 to Watson's $25000, and Brad's still in it... So it could be fun?!
... But of course, our "new computer overlord" has won!  H/t to Ken Jennings...

Epilogue:

Discuss Brian Christian's article... what is "human intelligence", anyway?

Discuss Fast Company article: http://www.fastcompany.com/node/1726969/print
... from one year ago
... "programmed delay"
... electromechanical hand?

Discuss Infoworld article: http://www.infoworld.com/print/151798

... "How IBM's Watson hammered its 'Jeopardy' foes"

- Explain why "Machine(s)" versus "Machine"

- Discuss latency as a "human" factor that "reflects" morality and wisdom (wow...)
... example in the "markets"
... in memory of John Boyd (OODA Loop is a feedback loop... but still a synchronous concept)
... in "life" ("Wait... count to ten... and then decide/act")

Edward Goldstick is a veteran of the high-tech, software, defense, and energy-technology worlds in the U.S. and France.

This article available online at:

http://www.theatlantic.com/technology/archive/2011/02/men-v-machine-s-the-rest-of-the-story/71381/