Recently I found this funny link.

From the text:

Computers mimic human reasoning by building on simple rules and statistical averages. Test your strategy against the computer in this rock-paper-scissors game illustrating basic artificial intelligence. Choose from two different modes: novice, where the computer learns to play from scratch, and veteran, where the computer pits over 200,000 rounds of previous experience against you.

Play it, and post your results.

First time, I played against the computer in ‘veteran’ mode.

What does it mean that I could beat the computer, but so narrow, and so many ties?

Second time, I played ‘novice’ mode. I lost terribly. Why?

Third time I succeeded to keep up against the novice. How did I do that? (I stopped at 27 rounds, because I liked the symmetry of the result…)

Things to think about: can a computer better play random than I can? Can I play better random than the average player? Can I guess the rule the ‘novice’ is using, or the average player? What is your experience?

Humans are terrible at being random.  This website has been recording data on humans vs its AI in RPS for a while; it has won almost 60% of the time:
http://www.essentially.net/rsp/index.jsp

Mingy Jongo - 22 February 2012 07:14 AM

Humans are terrible at being random.

Another piece of evidence against libertarian free will ...

I refuse to participate in this improper game that does not include either lizard or Spock. Fail.

OK, found a Website where you can play against different programs, or let compete different programs.
http://www.rpscontest.com/

I won when playing against a novice and lost when playing against a veteran. I then prepared 51 pieces of paper numbered 1,2, and 3: 1 for rock, 2 for paper, 3 for scissors. I took them out one by one from a bag and won both against the novice and the veteran; I won by a bigger margin against the novice, which was to be expected.

The computer learns but so do I. It probably wasn’t programmed to detect that I can help myself to play randomly. Tough luck, AI. 

The computer isn’t being random!!  That’s why it is learning!!!  It’s using algorithims. It’s using math and statistics to startegically predict opponents next moves!!
Random doesn’t involve predicting.

I played 100 rounds. Veteran.  I could easily begin to see the pattern that the computer was using to beat me.  I threw 10 consecutive rocks.  By the 3rd time the computer faithfully threw out paper every time.  That’s not random. It’s complicated math that most humans are incapable of doing.
26 wins 22 ties 52 losses.

VYAZMA - 22 February 2012 11:19 AM

The computer isn’t being random!!  That’s why it is learning!!!  It’s using algorithims. It’s using math and statistics to startegically predict opponents next moves!!
Random doesn’t involve predicting.

It should have learned to be random when playing with me. That’s why it lost.

2nd round. Veteran.
100 turns.
wins:33
ties:36
loss:31
Just make a pattern like left right 4 times. Then right 2 times in a row. then left right 4 times again then center. Repeat variations of this and slightly tweak one component every other series. Then every other second series etc…
(left-rock   center-paper   right-scissors)

George - 22 February 2012 11:43 AM

VYAZMA - 22 February 2012 11:19 AM

The computer isn’t being random!!  That’s why it is learning!!!  It’s using algorithims. It’s using math and statistics to startegically predict opponents next moves!!
Random doesn’t involve predicting.

It should have learned to be random when playing with me. That’s why it lost.

If was actually a bit funny to read its strategy when I knew I was playing randomly. It sounded like Dr. Phill, a sports commentator or an economist when I was reading its strategy thinking.

I played veteran.  I was leading by a slight margin until I almost reached 100 total, (when I had hoped to stop and declare victory), however the AI caught up and took a lead.  I kept playing determined to get a tie, which I did at 44-48-44.  I fear that if I had kept playing the algorythyms would have surpassed my ability to catch up again.

I played until I beat the veteran AI by one. Stopped at 50/48/49. Very close matchup. It was winning by about 10 points most of the time, but I was able to catch up somehow. I notice the AI doesn’t tend to throw 3 or more of the same in a row. Sometimes I’ll throw like 6 or more of the same in a row. Works on humans, but not so well on this AI.

dougsmith - 22 February 2012 07:22 AM

Mingy Jongo - 22 February 2012 07:14 AM

Humans are terrible at being random.

Another piece of evidence against libertarian free will ...

How so?

VYAZMA - 22 February 2012 11:25 AM

I played 100 rounds. Veteran.  I could easily begin to see the pattern that the computer was using to beat me.  I threw 10 consecutive rocks.  By the 3rd time the computer faithfully threw out paper every time.  That’s not random. It’s complicated math that most humans are incapable of doing.
26 wins 22 ties 52 losses.

That’s interesting. It doesn’t say anything about considering the current opponent’s play style when it’s thinking, although it totally should.