Tuesday, February 15, 2011

Reaction Time

"Watson" didn't have to push a button, a huge advantage when everyone knows the answer well before the question ends on "Jeopardy".

You could see it in real time, but I'd like to see data equivalent to what was used to determine the lock-out time for a false start in track when they shifted to electronic timing. I'd guess that "Watson" rung in at the same instant every time as it didn't have to worry about moving a finger or the delay as flesh meets button. After all, it could devote a processor to that task without any distractions from the main task at hand.

Indeed, this raises two questions about how "Jeopardy" determines when the board is open to ring in with an answer.

1) Does a human press a button, or is the end determined by an electronic sensor listening to the host's microphone? If the latter, a machine with a non-mechanical reaction time can learn to anticipate when the lights will come on.

2) Does the system lock out for a reasonable reaction time after the lights come on, the way they do for false starts in track? In track, extensive tests with world class sprinters showed that none had a reaction time less than 0.1 s, with the best around 0.12 s. Since visual responses are, reportedly, slower, does Jeopardy have a 0.2 s lockout for a valid response?

If it doesn't, "Watson" had a huge advantage.

4 comments:

Sherman Dorn said...

From "Science Friday" last week:

FLATOW: Does it have a mechanical buzzer, like a finger?

Mr. BAKER: That was a point of controversy, because IBM originally didn't have one. They were basically sending email. Watson was buzzing in electronically. And the "Jeopardy!" people said, hey, wait a minute, this isn't fair. You guys have to build a finger for that machine, and it has to press the physical buzzer just like the humans.

And IBM said: Wait a minute. You're turning - we built a brain, and you're turning it into a robot. This is like a Frankenstein monster. They said: You're grafting human limitations on to our machine. But they obeyed.

http://www.npr.org/2011/02/11/133686004/IBM-Computer-Faces-Off-Against-Jeopardy-Champs

Doctor Pion said...

Thanks for that pointer!

But reaction time is more than just mechanical movement, which can be much faster with a suitable solenoid and switch. The human limitation is photochemical-preprocess network-brain-nerves-muscle. Human vision is complicated and has limits that machine vision does not.

I was also irritated that it didn't have to actually write its final answer, and amused that I could guess "Chicago" along with both champions ... and it could not. That was a great question!

Stu said...

I thought the ability to mechanically actuate "the finger" (however IBM determined to best time the mechanism) was just part of the package of being a machine vice being human.

Just as not being "smart" enough to adapt its answer based on incorrect responses from the other participants is a disadvantage.

Doctor Pion said...

Ah, but it was sold as a contest of smarts, not reaction time. I'd like to see the data.