[Noisebridge-discuss] But rock, paper, scissors?!

[lordclaypool]

Well, this system isn't too difficult to figure out. They used a high speed
camera to predict future state estimations of the human hand based on prior
data (fractions of a second before the hand forms a rock, paper, or
scissor). What's better is that they then fed it into a learning algorithm
and kept repeating rock, paper, scissors examples to it, thus positively
reenforcing the learning algorithm. Once you let it lose, it can figure out
what the gesture will be very early. That gives this super fast moving hand
plenty of time to form it's decision. Human hands are slow compared to this
machine :) Kudos to the team to mechanically engineered that thing.

They've done great work, but it's hardly comparable to Watson.

Here's a link to their website and publications:
http://www.k2.t.u-tokyo.ac.jp/fusion/Janken/index-e.html

-Lord Clayp00L

On Wed, Jun 27, 2012 at 3:30 PM, Jake <jake at spaz.org> wrote:

>
> http://www.csmonitor.com/Science/2012/0627/Robot-beats-human-at-rock-paper-scissors.-Every-time
> .
>
> It was somewhat more unsettling when they started mastering trivia. But
> even when IBM's supercomputer, Watson, defeated Jeopardy! champion Ken
> Jennings last year, we meatbags could comfort ourselves by knowing that
> all the machine was really doing was plucking keywords and searching the
> encyclopedias stored in its four-terabyte hard disk.
>
> But rock, paper, scissors?!
>
> Apparently what most of us had assumed was an innocent game of chance, an
> equitable way of say, determining who gets to ride shotgun, is actually a
> computational problem that can be measured, analyzed, and, if your'e a
> robot with a high-speed camera and lightning-quick motors, mastered.
>
> A robot at the University of Tokyo's Ishikawa Oku Lab has done just that.
> Dubbed the "human-machine cooperation system"  "cooperation" being a
> robotics term for "humiliation"  this mechanical hand is able to beat a
> human opponent at rock paper scissors. Every time.
> _______________________________________________
> Noisebridge-discuss mailing list
> Noisebridge-discuss at lists.noisebridge.net
> https://www.noisebridge.net/mailman/listinfo/noisebridge-discuss
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.noisebridge.net/pipermail/noisebridge-discuss/attachments/20120627/58bb5601/attachment-0003.html>