Comments on: It’s a sweep!

By: Wayne

Wayne — Tue, 15 Mar 2016 00:06:18 +0000

Dick:
Thank you! (I can actually follow that…)
Now, I am more impressed!

By: Dick Eagleson

Dick Eagleson — Mon, 14 Mar 2016 23:32:00 +0000

From what I’ve read about the software being used by the supercomputer that beat Mr. Sedol, it is a generic machine-learning AI that is not optimized for the game of Go or even for playing games in general. That would seem to make this result even more significant than Deep Blue’s defeat of Gary Kasparov at chess nearly 20 years ago. Deep Blue’s software was entirely chess-centric and, if I recall correctly, even the hardware was not general-purpose. I believe each computing node included a hardware chess move generator.

By: Blair Ivey

Blair Ivey — Mon, 14 Mar 2016 06:44:41 +0000

It seems that as a species we have developed a new way of looking at the universe. The non-biological program keeps ‘its eyes on the prize’ and sacrifices short-term gain for long-term success. Biological entities (like humans) tend to be more concerned with short-term goals and try to make them work for long-term strategies. I would hope that military colleges take note of this at once both horrifying and fascinating development.

By: Wayne

Wayne — Mon, 14 Mar 2016 05:26:35 +0000

Garry–
Thanks for expanding on that.
I understand basic game-theory, just not how or what element’s are used in the program.
(not familiar with the actual game either– a drawback for me on understanding.)

Is this designed to learn from repetitive play or is it following strictly deterministic processes or both? (I realize I’m not using these terms, precisely.)

–I’m currently (trying!) getting a handle on P Vs. NP problems in computation, which is not directly related but I could see where it would be a factor of sorts, “depending.”

As I understand it– Chess for example, can be “beaten” with simple brute force calculation, but in certain set-ups of the pieces, a program would miss (not “recognize that..) moves-in-the-future not yet made would cause it to lose, where-as any novice chess-player can overtly “see” it taking place ahead of time..

I don’t have a grasp of the conditional-rules a program would need to be able to calculate complex, multiple “if-then” linear or non-linear solution’s.

Sorry– totally babbling & not adequate describing what I think I might mean. (HAR)

Tangent– I’m a big fan of “emergent-order,” where-in seemingly chaotic situations, can expand out to “infinity” but always contract back down to solvable situations and predictable solutions.
— I’m thinking certain Fractal sets and or Penrose tiling problems & solutions… (but again, my eyes are bigger than my brain on all this.)
–fascinating stuff– nonetheless!!
Haven’t spent enough time on modern AI concepts to even form my questions clearly!!
— currently wrapped up in Leonard Susskind’s fascinating presentation of “Boltzmann & The Arrow of Time” part 1 of 3, as it relates to entropy & time-reversible process’s as they apply or don’t, to Cosmology… too much on my plate! HAR!

(thanks nonetheless!)

By: Garry

Garry — Mon, 14 Mar 2016 04:29:59 +0000

My interpretation is that the algorithm made decisions that looked much different from what a human would decide. No matter how calculating a human player is, in some cases he lets intuition dictate his decisions. In many realms, this is what makes humans superior to any kind of AI (as it exists now).

When we reduce the universe to a relatively simple game, however, computational power has advanced to the point where it can defeat even the most skilled human players; moves that “look” wrong can cloud our judgments and make us reject them too early.

I apologize for not having a link, but I read somewhere that Lee defeated the computer in the 4th game. I look forward to reading about implications in the field of human vs. machine learning.

By: Wayne

Wayne — Mon, 14 Mar 2016 03:14:04 +0000

“….The algorithm seems to be holding back its power. Sometimes it plays moves that lose material because it is seeking simply to maximise its probability of reaching winning positions, rather than — as human players tend to do — maximise territorial gains…..”

This sounds rather anthropomorphic to me– “it’s holding back it’s power.”
Correct me if I’m off-base, does it not “only do what it has been programmed to do?”

— Or am I missing the point & it’s more into the AI machine-learning realm?
(I can write a little HTML & troubleshoot my computer, darn well, but not familiar with modern programming much past BASIC & COBOL. That dates me eh!)
— Am familiar with the basics of Game Theory but not in the computer-applied realm.)