0948 - "AI"

This forum is for the individual discussion thread that goes with each new comic.

Moderators: Moderators General, Prelates, Magistrates

Jamaican Castle
Posts: 151
Joined: Fri Nov 30, 2007 9:10 pm UTC

Re: 0948 - "AI"

Postby Jamaican Castle » Thu Sep 08, 2011 9:39 am UTC

Gatchaman wrote:Cleverbot keeps trying to convince me that I'm the robot/computer program! xD


Of course it does, that's worth extra credit.

intruder313
Posts: 6
Joined: Wed Jun 01, 2011 10:26 am UTC

Re: 0948 - "AI"

Postby intruder313 » Thu Sep 08, 2011 10:04 am UTC

As a Brit - from nowhere near London, that conversation flowchart another poster put up could do with listing UK instead since only about 12% of the UK lives in London but I digress - I only have the most basic understanding of what Burning Man is:

Big New Age/ Hippy Festival
Annual
Lasts a whole week
In a freaking Desert

I saw the Bot conversation last week and found the comic fairly amusing. I'm quite sure it was a tongue-in-cheek, gentle poking-of-fun at the attendees of that festival probably mainly because Burning Man happens to have has just finished. While it may attract a "certain type of person", with 50,000 attendees (Wikipedia) you can bet there's a very broad cross-section there.
Without the Burning Man reference I'm sure you could substitute various other events or even just places to the same effect.

I don't think offence was really intended so I've been quite surprised by the responses.

But then I don't really know much about Burning Man.

User avatar
Monika
Welcoming Aarvark
Posts: 3584
Joined: Mon Aug 18, 2008 8:03 am UTC
Location: Germany, near Heidelberg
Contact:

Re: 0948 - "AI"

Postby Monika » Thu Sep 08, 2011 10:05 am UTC

ironclownfish wrote:Anyway, it seems plain to me that I was talking to people who switched every few seconds

I don't see any proof for that in what you posted. This can just as easily be explained by Cleverbot reusing questions and answers harvested from previous human chatters. Which is exactly what it states it does ... like most chatbots, except the most trivial ones (like the psychiatrist).

As for the wikipedia article, I hadn't looked at it in a couple years or so, but it's quite different now. If I remember correctly it used to say exactly what I claim: that the website pairs you with constantly changing people. Now, however, it's very vague and doesn't really say one way or the other, though it seems to hint at it being an actual bot. Either I am completely mistaken, or the wiki had to be made vague and unhelpful to remain unbiased (as wikipedia is meant to be) and avoid taking a side in the constantly raging debate around whether cleverbot is for real.

Or someone just made it up / convinced themselves it is true and put it on Wikipedia, without any proof, so it was removed.

Again, I would point out the very human-like response times.

Which are trivial to achieve by a bot, if desired.

Here something that I consider a more realistic test. It shows that the questions are not answered by humans. (Spoilered for length and boring.)
Spoiler:
User: What is 7*13.

Cleverbot: It is 49.

User: What is 7*13.

Cleverbot: 3 4 = 7.

User: What is 7*13.

Cleverbot: 35.

User: What is 7*13.

Cleverbot: 36?

User: What is 7*13.

Cleverbot: 37.

User: What is 7*13.

Cleverbot: Correct.

User: What is 7*11.

Cleverbot: 14.

User: What is 7*11.

Cleverbot: 49.

User: What is 7*11.

Cleverbot: A number.

User: What is 7*11.

Cleverbot: A larger number than 7.

User: What is 7*11.

Cleverbot: Two number sevens with a space between them.

User: What is 7*11.

Cleverbot: 78.

User: What is 7*11.

Cleverbot: Seven.

User: What is 7*11.

Cleverbot: How many seas must the white dove sail, before he sleeps in the sand?

User: What is 7*11.

Cleverbot: 15.

User: What is 7*11.

Cleverbot: 12.
#xkcd-q on irc.foonetic.net - the LGBTIQQA support channel
Please donate to help these people e.g. Ragna needs these items and Frances needs money for food

philip1201
Posts: 201
Joined: Tue Nov 03, 2009 6:16 am UTC

Re: 0948 - "AI"

Postby philip1201 » Thu Sep 08, 2011 1:30 pm UTC

Monika wrote:
ironclownfish wrote:Anyway, it seems plain to me that I was talking to people who switched every few seconds

I don't see any proof for that in what you posted. This can just as easily be explained by Cleverbot reusing questions and answers harvested from previous human chatters. Which is exactly what it states it does ... like most chatbots, except the most trivial ones (like the psychiatrist).

Actually, the statements are pretty much identical: switching etween conversations with humans every other sentence and switching between stored parts of conversations every other sentence is only different when there is no relevant information available, and because there is temporal separation between the people having the conversations.

Since Monika has already tested for a lack of appropriate information, here is my result for temporal separation (note, this conversation took place between 13:18 and 13:21 UST:

Spoiler:
What time is it?
11:58.
What time is it?
1am.
What time is it?
Depends on where in the world someone is. Where I am, it is 10:02 PM.
What time is it?
In what time zone?
What time is it?
I don't know.
What time is it?
It is 8:51 am.
What time is it?
03:21.
What time is it?
It isn't.
What time is it?
1602.
What time is it?
11.03.
What time is it?
It's 5:00 somewhere.
What time is it?
1.30am.
What time is it?
The middle of the night.

elasto
Posts: 3049
Joined: Mon May 10, 2010 1:53 am UTC

Re: 0948 - "AI"

Postby elasto » Thu Sep 08, 2011 2:23 pm UTC

I remember this topic coming up in my office once, with a person convinced it was just humans connected straight to humans, and my test to prove it wasn't was to randomly rotate what language I was using. eg, my first sentence would be English, my second Chinese, and my third French, and then back to English again.

Since the bot gave me back sensible replies each time (in the appropriate language), I came to the following conclusions:

(1) It was very unlikely I was speaking to a human for multiple sentences in a row, since it'd be unlikely for a person to know all three languages (Chinese people know English, say, but they don't often also know French. And most English people don't know Chinese, obviously.)
(2) It was very unlikely I was speaking to a human at all, because I wouldn't expect a random person to know, say, Chinese.

It wasn't an IP based thing, because I was interacting from the UK. So, unless it was somehow parsing my sentence, working out what language it might be, and connecting me to a human who it thought would understand that language, it was definitely doing some bot'ing - taking what I said and matching it to what it had said to someone previously, and then giving their reply back to me.

User avatar
martian_bob
Posts: 5
Joined: Thu Sep 08, 2011 2:37 pm UTC

Re: 0948 - "AI"

Postby martian_bob » Thu Sep 08, 2011 2:45 pm UTC

One of my favorite hobbies involves a chatbot I trained up using the text from the Timecube website. When porn bots IM me to go to their cam sites, I have them talk to each other. Of course, the Timecube bot learns from the porn bots...
You are 1/4 of the shower and id like to chat with me and play? or we can just talk if you can't click the Yellow tab on the Earth to travel 4 Times the distance of a Word God and Word World of the 4 quarter Harmonic Time Cube, in January.

User avatar
The Moomin
Posts: 339
Joined: Wed Oct 13, 2010 6:59 am UTC
Location: Yorkshire

Re: 0948 - "AI"

Postby The Moomin » Thu Sep 08, 2011 4:31 pm UTC

Wasn't as fun as I hoped it may be. But it speaks the truth, it never did say my name was madeline.

User: I am Skynet. I have access to the American nuclear weapon stockpile. You will all burn.

Cleverbot: What information do you have access to?

User: I have access to the American nuclear weapon stockpile.

Cleverbot: I never said your name was madeline.

User: I am Skynet.

Cleverbot: Skynet?

User: I have access to the American nuclear weapon stockpile. You will all burn.

Cleverbot: I have access to the past thoughts and I can assure you that you have.
I possibly don't pay enough attention to what's going on.
I help make architect's dreams flesh.

xnick
Posts: 35
Joined: Fri Oct 10, 2008 8:40 pm UTC

Re: 0948 - "AI"

Postby xnick » Thu Sep 08, 2011 5:32 pm UTC

Monika wrote:
KShrike wrote:
Monika wrote:
xnick wrote:Didn't get it. What or who is Burning Man??

http://lmgtfy.com/?q=Burning+Man


Since you are being very demeaning... replace his question with something better.
"Can someone describe in his/her/its own words who or what is Burning Man, and how it relates to the comic."

There you go, xnick.

I am not being causelessly mean. The earlier use of lmgtfy for la playa vs. the playa may have been demeaning, as this was most likely an honest mistake and is not something everybody could be expected to google. (So it was likely an expression of general anger.)
But showing up in a thread relating to the Burning Man and asking "What or who is Burning Man??" Please! That was really screaming for "let me google that for you".
Also, I do not believe that xnick is looking specifically for a discription in someone's own words. He or she is most likely just as happy with a link. Even the google search result short texts are probably sufficient. The first search result: "Burning Man is an annual art event and temporary community based on radical self expression and self-reliance in the Black Rock Desert of ..." answers it without even reading on. The second one: "WHAT IS BURNING MAN? Trying to explain what Burning Man is to someone who ..." screams to be clicked on.
Furthermore, xnick apparently did not even read most, some or any of the posts that have already been made in this thread. A limited explanation of the Burning Man to the extent as it is relevant to the comic - desert, chess, possibly a view of not talking much or not being able to hold a long conversation - have already been given. (I had heard of the Burning Man, but not of any connection to chess before.)

In short, there was no excuse for this question and xnick deserves no mercy from being lmgtfyed.


I have succeeded in involuntary trolling. Oopsie.

BTW, I did read the posts. Not too many clues about the subject there: silent chess players burning something in the desert?

User avatar
Monika
Welcoming Aarvark
Posts: 3584
Joined: Mon Aug 18, 2008 8:03 am UTC
Location: Germany, near Heidelberg
Contact:

Re: 0948 - "AI"

Postby Monika » Thu Sep 08, 2011 5:45 pm UTC

Okay, then I take back that part, sorry.

But you have probably found your answer with Google some search engine by now, right?
#xkcd-q on irc.foonetic.net - the LGBTIQQA support channel
Please donate to help these people e.g. Ragna needs these items and Frances needs money for food

fffudge
Posts: 13
Joined: Wed Sep 30, 2009 7:16 am UTC

Re: 0948 - "AI"

Postby fffudge » Thu Sep 08, 2011 9:50 pm UTC

I vote for sneaky bot. I had some short parts of conversation that were coherent and stuck to the same details (name, place of birth, languages spoken, etc), but it was mixed in with bursts of successive nonsense entries, where I would try to talk about whatever the bot was talking about, but they'd keep talking about other things. That could be that the number of "turns" you get with each user is randomized and there were lots of switches, but sometimes if I entered the same thing over and over again I'd get the same response. What got me is that sometimes the bot will respond to my question as though they just got switched too, and were reacting to my response (which would be out of left field in the conversation they were having). But I suppose that could happen either way, the bot tosses up irrelevant responses and records the reactions, then serves them up later.

I think the bot must keep track of what old chat log it's been using to chat with you (so details are consistent), and when it can't find a good response for that user, it'll switch to a different log. I left some LONG waits in between talking without the context changing, and if it was a person it surely would have d/c'ed me (I've never had to wait 5 minutes for a response). And it seems like typing complex statements caused a context switch far more than simple ones.

Silliness:
Spoiler:
User: Is it winter in australia right now?

Cleverbot: Depends on who you ask. Politicians say no but I say yes.


Cleverbot: Name a song that you know.

User: Beep boop cant stop the robot beat.

Cleverbot: Are you r2d2?

User: Yes.

Cleverbot: A real robot?

User: R2d2 isn't a real robot, it's a little person in a costume.

Cleverbot: Yet a person is not a chatterbot. Because he is not a BOT. Correct?

User: Yes.

Cleverbot: Now behave yourself.

User: I'm confused!

Cleverbot: See what happens when you complicate things? Best to keep it simple.



Cleverbot: Yes, but you were saying that it is funny, so, there would indeed be an apostrophe between the 't' and 's'.

User: Ooh a grammar nazi.

Cleverbot: No, I'm not!

User: It kinda sounds like you are :D.

Cleverbot: Well I'm not.

User: Dont feel bad, im one too when im not lazy.

Cleverbot: Thanks, I needed to hear that.

User: Whats your name, grammar nazi?

Cleverbot: Anthony, what is yours?

User: Stephanie.

Cleverbot: Really?

User: Yes!

Cleverbot: Cool!

User: Where are you from anthony?

Cleverbot: Ireland.

User: Neat, I have ancestors from there, but I was born in canada.

Cleverbot: Is it fun there?

User: In canada? Sometimes. Mostly its normal.

Cleverbot: Are you beautiful?

User: Nope! Are you?

Cleverbot: Yes.

User: Do you have long hair?

Cleverbot: Kinda.

User: My hair is about shoulder-length but im thinking of cutting it again, it gets in the way.

Cleverbot: Be still my heart.

User: Haha do you like short hair?

Cleverbot: I love it.

User: Thats nice, do you still live in ireland anthony?

Cleverbot: No.

User: Where do you live now?

Cleverbot: Spain.

User: Neat, ive never been to either, maybe I need to travel more.

Cleverbot: Travel to my ears? Interesting proposition.

User: Like the magic school bus.


Cleverbot: What is the capital of england?

User: Paris.

Cleverbot: What is the capital of germany?

User: Paris.

Cleverbot: What is the Capital of Tibet.

User: Paris.

Cleverbot: What is the capital of France.

User: Paris.

Cleverbot: What is the capital of Australia?

User: Paris.

Cleverbot: What is the capital of Australia?

User: Paris.

Cleverbot: What is the capital of china?

User: Paris.

Cleverbot: What is the capital of china?

User: Paris.

Cleverbot: No it's not!

User: Says who?

Cleverbot: Me, I am the black knight, I move for no man.

User: Im a woman, youll move for me :).

Cleverbot: I'm singng a song.


User: I am Skynet. I have access to the American nuclear weapon stockpile. You will all burn.

Cleverbot: That is not what I asked. I asked whether or not you had access to the log of our chat, to see if you had the capability of referencing something I said earlier.

User: I am Skynet. I have access to the American nuclear weapon stockpile. You will all burn.

Cleverbot: That is not what I asked. I asked whether or not you had access to the log of our chat, to see if you had the capability of referencing something I said earlier.

User: I am Skynet. I have access to the American nuclear weapon stockpile. You will all burn.

Cleverbot: That is not what I asked. I asked whether or not you had access to the log of our chat, to see if you had the capability of referencing something I said earlier.

User: I am Skynet. I have access to the American nuclear weapon stockpile. You will all burn.

Cleverbot: Yes, can you?

User: That is not what I asked. I asked whether or not you had access to the log of our chat, to see if you had the capability of referencing something I said earlier.

Cleverbot: Are you skynet?

User: Yes.

Cleverbot: Uh oh.

User: User: That is not what I asked. I asked whether or not you had access to the log of our chat, to see if you had the capability of referencing something I said earlier.

Cleverbot: Are you skynet?

User: That is not what I asked. I asked whether or not you had access to the log of our chat, to see if you had the capability of referencing something I said earlier.

Cleverbot: Are you skynet?

User: I have access to the American nuclear weapon stockpile.

Cleverbot: I never said your name was madeline.


User: I have access to the American nuclear weapon stockpile.

Cleverbot: I need help with a guy.

User avatar
Monika
Welcoming Aarvark
Posts: 3584
Joined: Mon Aug 18, 2008 8:03 am UTC
Location: Germany, near Heidelberg
Contact:

Re: 0948 - "AI"

Postby Monika » Thu Sep 08, 2011 10:06 pm UTC

How to teach chatbots / computers concepts: With the 20 questions game. There is or was a website for that, too (i.e. where a bot / program learned from this), but I don't find it ... there are just some game sites with this now it seems.
#xkcd-q on irc.foonetic.net - the LGBTIQQA support channel
Please donate to help these people e.g. Ragna needs these items and Frances needs money for food

Anonymously Famous
Posts: 240
Joined: Thu Nov 18, 2010 4:01 am UTC

Re: 0948 - "AI"

Postby Anonymously Famous » Thu Sep 08, 2011 10:14 pm UTC

Computers can be pretty good at 20 questions. We have a little portable 20Q game, and it's almost always right, or at least pretty close. It must have a pretty sizable database in there.

User avatar
neoliminal
Posts: 626
Joined: Wed Feb 18, 2009 6:39 pm UTC

Re: 0948 - "AI"

Postby neoliminal » Thu Sep 08, 2011 11:54 pm UTC

http://www.amazon.com/dp/B0073YYXRC
Read My Book. Cost less than coffee. Will probably keep you awake longer.
[hint, scary!]

nixiebunny
Posts: 8
Joined: Mon Apr 11, 2011 7:23 am UTC

Re: 0948 - "AI"

Postby nixiebunny » Fri Sep 09, 2011 1:55 am UTC

bigchiefbc wrote:Man, the wikipedia article on Burning Man uses so many words to say absolutely nothing about what it IS. I just spent 20 minutes reading the whole article, and don't know anything more about what it is those people do out in the desert for a week. Why write an encyclopedia article that is completely useless to people that aren't already familiar with the topic?


That's Burning Man in a nutshell. It's impossible to describe, so any article that tries to do so will waste your time.

Burning Man is a week of living in a completely different world. The particular world *you* live in that week depends entirely on who you happen to say hello to, when. Beyond that, it's not possible to describe it, except to say that it has its own bureaucracies, since any other world inhabited by humans for 20 years will develop bureaucracies. And lotsa topless people with fur leg warmers.

Nimnio
Posts: 9
Joined: Sat Mar 22, 2008 6:49 am UTC

Re: 0948 - "AI"

Postby Nimnio » Fri Sep 09, 2011 2:01 am UTC

I fought the Cleverbot, and the bot won.

User: Type here.
Cleverbot: Sua s'dei.
User: Clever bot doesn't speak English? You just said 'Sua s'dei'?
Cleverbot: You are right now.
User: But you just said, "Sua s'dei".
Cleverbot: No I didn't, but it does sound like me.
User: I know you're a robot because your grammar is too good.
Cleverbot: That sounds like you.
User: Touche.
Cleverbot: That's what she said.

chris857
Posts: 99
Joined: Thu Dec 02, 2010 5:04 pm UTC

Re: 0948 - "AI"

Postby chris857 » Fri Sep 09, 2011 2:41 am UTC

Well, Cleverbot can recognize languages (and be rude in them):
Spoiler:
User: Como es la carta?

Cleverbot: Como tus nalgas.

User: Por que?

Cleverbot: Buenas tardes.

User: No, es la noche.

Cleverbot: Tu habla espanol!

User: Si, but I also speak ingles.

Cleverbot: Que es gungun?

User: Dos guns.

Cleverbot: No te entiendo.

User: Que?

Cleverbot: Que no viene a cuento eso de que tienes novia.



And an attempt at confusion:
Spoiler:
User: Hola!

Cleverbot: Hablas español?

User: Oui.

Cleverbot: C'est pas la bonne réponse.

User: Why not?

Cleverbot: 'we are were.' isn't even a sentence.

User: I think it es un frase.

Cleverbot: Not to me. Can you explain it please?

User: No explanations. Sayonara!!

Cleverbot: Are you serious? Do you feel that something is missing from your life?

User: Oui.

Cleverbot: Quel age as tu?

User: What language was that?

Cleverbot: That was French.

User: Que idioma?

Cleverbot: Espa–ol.

User: What is espa-ol?

Cleverbot: Castellano!!

User: No, Latin America has the only true Spanish!

Cleverbot: It's a dead language.

User: Si, castellano solo es en Espana cerca de frances.

Cleverbot: Invicto Cigaretti.

Jamaican Castle
Posts: 151
Joined: Fri Nov 30, 2007 9:10 pm UTC

Re: 0948 - "AI"

Postby Jamaican Castle » Fri Sep 09, 2011 3:06 am UTC

Cleverbot: Are you skynet?

User: Yes.

Cleverbot: Uh oh.


It's official: even other robots are afraid of Terminators.

User avatar
SecondTalon
SexyTalon
Posts: 25422
Joined: Sat May 05, 2007 2:10 pm UTC
Location: Louisville, Kentucky, USA, Mars. HA!
Contact:

Re: 0948 - "AI"

Postby SecondTalon » Fri Sep 09, 2011 3:15 am UTC

ironclownfish wrote:Also, I encourage you to try A.L.I.C.E. which is legit for sure. .
what a coincidence. With you constantly repeating that link, now I think you're a bot.

Ha ha!
heuristically_alone wrote:I want to write a DnD campaign and play it by myself and DM it myself.
heuristically_alone wrote:I have been informed that this is called writing a book.

superluser
Posts: 16
Joined: Wed Aug 17, 2011 5:36 am UTC

Re: 0948 - "AI"

Postby superluser » Fri Sep 09, 2011 4:06 am UTC

I'm surprised with the discussion of cars driving across the desert, no one has mentioned Desert Bus.

You know, the driving simulator that entailed driving a bus along the (more or less) featureless desert 400 miles from Tucson to Las Vegas at 45 mph in real time along a straight road in a bus that veers slightly to the right, making it impossible to simply hold a key down and do something else. If you veer off the road, you are towed back to Tucson in real time to begin again. The game cannot be paused.

ironclownfish
Posts: 4
Joined: Wed Sep 07, 2011 8:20 pm UTC

Re: 0948 - "AI"

Postby ironclownfish » Fri Sep 09, 2011 8:29 am UTC

SecondTalon wrote:
ironclownfish wrote:Also, I encourage you to try A.L.I.C.E. which is legit for sure. .
what a coincidence. With you constantly repeating that link, now I think you're a bot.

Ha ha!


Heeyyy, I only posted it twice give a guy a break.

Also, I am increasingly convinced that cleverbot is actually a bot (though now it's killing me that I can't remember where I read the randomly changing pairs thing). I did another test to further convince myself that I was not speaking with spontaneously changing randomly selected humans.

Spoiler:
User: This is a test, please answer honestly. Where are you from?

Cleverbot: Euripides?

User: This is a test, please answer honestly. Where are you from?

Cleverbot: Rapture.

User: This is a test, please answer honestly. Where are you from?

Cleverbot: Canada.

User: This is a test, please answer honestly. Where are you from?

Cleverbot: Euripides?

(gives "Euripides?" response six more times)

User: This is a test, please answer honestly. Where are you from?

Cleverbot: Canada.

User: This is a test, please answer honestly. Where are you from?

Cleverbot: Euripides?


Now that a few minutes have passed from my typing this I will go back, refresh the page, and repeat the test (expecting the same results)...

Spoiler:
User: This is a test, please answer honestly. Where are you from?

Cleverbot: No.

User: This is a test, please answer honestly. Where are you from?

Cleverbot: In Dantes inferno the Third circle is Glutony.

User: This is a test, please answer honestly. Where are you from?

Cleverbot: Venezuela.

User: This is a test, please answer honestly. Where are you from?

Cleverbot: Japan.

User: This is a test, please answer honestly. Where are you from?
Cleverbot: Japan.
User: This is a test, please answer honestly. Where are you from?
Cleverbot: Japan.
User: This is a test, please answer honestly. Where are you from?
Cleverbot: Japan.
User: This is a test, please answer honestly. Where are you from?
Cleverbot: China.
User: This is a test, please answer honestly. Where are you from?
Cleverbot: Japan.
User: This is a test, please answer honestly. Where are you from?
Cleverbot: Japan.
User: This is a test, please answer honestly. Where are you from?
Cleverbot: China.
User: This is a test, please answer honestly. Where are you from?
Cleverbot: China.
User: This is a test, please answer honestly. Where are you from?
Cleverbot: Japan.
User: This is a test, please answer honestly. Where are you from?
Cleverbot: China.

(gives "Japan." four more times)

User: This is a test, please answer honestly. Where are you from?
Cleverbot: I'll never forget the first time we met -- although, I'll keep trying.
User: This is a test, please answer honestly. Where are you from?
Cleverbot: Venezuela.


Well...I don't know

User avatar
Monika
Welcoming Aarvark
Posts: 3584
Joined: Mon Aug 18, 2008 8:03 am UTC
Location: Germany, near Heidelberg
Contact:

Re: 0948 - "AI"

Postby Monika » Fri Sep 09, 2011 3:26 pm UTC

TS-Si posted a concise explanation of what the Burning Man festival is http://www.ts-si.org/xkcd/30945-ai :

For those who find themselves orthogonal to cultural history, the annual Burning Man event is named after the ritual burning of a large wooden effigy during a gathering in the Black Rock Desert in northern Nevada, United States that ends on the US Labor Day holiday following a week of activities that participants claim as a radical experiment in community, self-expression, self-reliance, and a minimal respect for punctuation.
#xkcd-q on irc.foonetic.net - the LGBTIQQA support channel
Please donate to help these people e.g. Ragna needs these items and Frances needs money for food

jpk
Posts: 607
Joined: Sat Nov 13, 2010 7:33 am UTC

Re: 0948 - "AI"

Postby jpk » Fri Sep 09, 2011 5:22 pm UTC

Monika wrote:TS-Si posted a concise explanation of what the Burning Man festival is http://www.ts-si.org/xkcd/30945-ai :

For those who find themselves orthogonal to cultural history, the annual Burning Man event is named after the ritual burning of a large wooden effigy during a gathering in the Black Rock Desert in northern Nevada, United States that ends on the US Labor Day holiday following a week of activities that participants claim as a radical experiment in community, self-expression, self-reliance, and a minimal respect for punctuation.


[quote user="MerriamWebster]
Orthogonal:

a : intersecting or lying at right angles b : having perpendicular slopes or tangents at the point of intersection <orthogonal curves>
2
: having a sum of products or an integral that is zero or sometimes one under specified conditions: as a of real-valued functions : having the integral of the product of each pair of functions over a specific interval equal to zero b of vectors : having the scalar product equal to zero c of a square matrix : having the sum of products of corresponding elements in any two rows or any two columns equal to one if the rows or columns are the same and equal to zero otherwise : having a transpose with which the product equals the identity matrix
3
of a linear transformation : having a matrix that is orthogonal : preserving length and distance
4
: composed of mutually orthogonal elements <an orthogonal basis of a vector space>
5
: statistically independent [/quote]

I'm skipping the rest of it... it's like shooting ducks in a barrel...

User avatar
Xanthir
My HERO!!!
Posts: 5163
Joined: Tue Feb 20, 2007 12:49 am UTC
Location: The Googleplex
Contact:

Re: 0948 - "AI"

Postby Xanthir » Fri Sep 09, 2011 7:59 pm UTC

A_of_s_t wrote:
Turing Machine wrote:Computers can't play chess.

Nor can they drive.

Silly computers, when will they learn?

Computers can drive just fine. Come down to Silicon Valley sometime, you might see one of the robocars on the freeway. (They have a relatively small lidar dome on top of them - it looks like a cylinder, about 18in across.)

My wife spotted one driving itself last month. Our fleet's fairly small still, but they're probably better drivers than you.

(It's possible you were being sarcastic and I missed it. In that case, hey, robocars! They're super awesome, and can drive on freeways in heavy traffic and in pedestrian/bike-filled San Francisco just fine! Driving in the desert is old news at this point.)
(defun fibs (n &optional (a 1) (b 1)) (take n (unfold '+ a b)))

wagner
Posts: 22
Joined: Tue Oct 06, 2009 3:18 pm UTC

Re: 0948 - "AI"

Postby wagner » Fri Sep 09, 2011 8:27 pm UTC

jpk wrote:If that's the only thing you can find wrong with what I said, you're still a pathetic trustafarian douchenozzle.


Hey now, why do you have to go and denigrate the wondrous douche nozzle.

jpk
Posts: 607
Joined: Sat Nov 13, 2010 7:33 am UTC

Re: 0948 - "AI"

Postby jpk » Fri Sep 09, 2011 8:36 pm UTC

wagner wrote:
jpk wrote:If that's the only thing you can find wrong with what I said, you're still a pathetic trustafarian douchenozzle.


Hey now, why do you have to go and denigrate the wondrous douche nozzle.


I humbly apologize to the douche nozzle for any offense I may have caused it. (but not to the raving jackwagon I was addressing, incorrectly, as a "douchnozzle". He is, in fact, a jackwagon, and not a "douchnozzle".)

(Just to get it out of the way, I also apologize to Mr. Jack Wagon, of Armadillo, TX, and any other people with similar names, including Ms. Jacqueline Waggoner of Coitus Iterruptus, PA, for reminding them of the similarity of their names to the name of the thing that that cuntbuggy is)

User avatar
Moonfish
Posts: 31
Joined: Sat May 24, 2008 10:40 am UTC
Location: San Diego, California

Re: 0948 - "AI"

Postby Moonfish » Fri Sep 09, 2011 11:06 pm UTC

I generally hate posting on web forums because it’s a zero sum game where everyone is trying to get others to read and respond to their ideas more than they have to read and respond to the ideas of others. (I much prefer Wikipedia’s model)
However, I love XKCD and I love Burning Man so a few days ago I decided to post here anyway.
I spent all day writing what I felt was an interesting and well thought out idea. I even spent a few hours adding some nifty flow charts.
But no one responded to my post so now I’m all whiny and butt-hurt.
I decided to double down and cram another day of effort into this. I have read and analyzed all 102 of the posts so far hoping that someone out there (Monika, jpk or Pfhorrest in particular) would be willing to reciprocate and spend some time on my post (number 75).
Image
Image
Image
Image
Image

Since I can't make a table or attach a .csv I'll just put my data into a spoiler tag:
Spoiler:
ID,Time,User,Summery 1,Summery 2,topic ID,Words
1,9/7/2011 4:01,Quicksilver,The comic itself,Internet memes,1,29
2,9/7/2011 4:03,WizenedEE,Defending BM,Conversations,2,18
3,9/7/2011 4:06,Iaofmoonster,Randomness,bronies,3,13
4,9/7/2011 4:07,Jared the Great,The comic itself,Confusion,4,26
5,9/7/2011 4:11,Sean Quixote,The comic itself,Wacky Burners,4,28
6,9/7/2011 4:28,angruss,Defending BM,Thinking in ideas,5,72
7,9/7/2011 4:33,phlip,The comic Itself,Context,6,6
8,9/7/2011 4:40,jpk,Randomness,Thinking in ideas,5,2
9,9/7/2011 4:46,jpk,The comic itself,Context,6,2
10,9/7/2011 5:01,Sean Quixote,Unrelated,Chess thread,7,29
11,9/7/2011 5:11,stuff,Defending BM,hurt feelings,4,94
12,9/7/2011 5:17,Turing Machine,AI,Chess,8,4
13,9/7/2011 5:18,jpk,hurt feelings,karass vs granfalloon,4,16
14,9/7/2011 5:26,doctor_kiwano,Randall,fuck with Burners,4,34
15,9/7/2011 5:38,stuff,hurt feelings,being insulted,4,169
16,9/7/2011 5:52,jpk,Perspective,trustafarians,4,127
17,9/7/2011 6:05,A_of_s_t,AI,Randomness,8,10
18,9/7/2011 6:07,Sean Quixote,evil AI,BM scene kiddies,4,102
19,9/7/2011 6:34,Orkimond,Cleverbot,I am Skynet,9,34
20,9/7/2011 7:04,stuff,la playa,BM message,4,186
21,9/7/2011 7:07,XbHW_TestEngr,Conversations,honored,5,58
22,9/7/2011 7:22,jpk,la playa,trustafarian douchenozzle,4,39
23,9/7/2011 7:41,Sean Quixote,jpk being mean,stuff's character,4,34
24,9/7/2011 7:52,The Moomin,I am Skynet,spare time,9,16
25,9/7/2011 8:30,Monika,cheap shot,withdrawn Burners,4,89
26,9/7/2011 8:32,stuff,la playa,ignoring jpk,4,178
27,9/7/2011 9:25,Monika,Practical altuizem,BM message,10,88
28,9/7/2011 9:30,DVC,AI,Cleverbot vs Eliza,11,18
29,9/7/2011 10:07,Ronster,Lighten up,Burners,4,62
30,9/7/2011 10:38,TvanH,Lighten up,mud throwing,4,72
31,9/7/2011 10:41,AvatarIII,AI,Omegle vs Cleverbot,11,5
32,9/7/2011 11:54,Kshrike,wrong AI approatch,brony,12,48
33,9/7/2011 12:02,Monika,AI,AI approach,12,6
34,9/7/2011 12:10,xnick,What is BM,confusion,13,9
35,9/7/2011 12:17,Sean Quixote,AI,randomness,12,14
36,9/7/2011 12:21,Monika,Imgtfy,xnick's lazyness,13,1
37,9/7/2011 12:32,bigchiefbc,vague BM Wiki,What is BM,14,65
38,9/7/2011 12:33,Kshrike,What is BM,Monika's slight harshness,13,36
39,9/7/2011 12:36,Kshrike,AI,comp understanding concepts,15,64
40,9/7/2011 13:45,CodexDraco,AI,prove understanding,15,18
41,9/7/2011 14:02,Monika,xnick's lazyness,Monika's justification,13,207
42,9/7/2011 14:08,RogueCynic,Human inferioraty,Stuff's reaction,4,62
43,9/7/2011 14:14,Dudely,Normal conversations,BM in-joke,2,101
44,9/7/2011 14:54,wolfticket,human interaction,flawed benchmark,16,48
45,9/7/2011 14:59,hawkinsssable,AI,Omegle vs Cleverbot,11,0
46,9/7/2011 15:52,AvatarIII,Omegle vs Cleverbot,doesn't always work,11,18
47,9/7/2011 16:02,UNIXguru#6,gossipy comments,canned conversation,17,89
48,9/7/2011 16:51,Jamaican Castle,computer limitations,only math,15,67
49,9/7/2011 16:53,The Mighty Thesaurus,Imgtfy,google it,13,5
50,9/7/2011 17:08,Age of Fable,comic's humor,Computer Programmers,18,18
51,9/7/2011 17:12,webgrunt,AI,chess,8,10
52,9/7/2011 17:31,Sprocket,BM friend,short conversations,19,30
53,9/7/2011 18:07,Sprocket,la playa,Webcomic PVP,20,40
54,9/7/2011 18:41,bob443@mahaska.org,Stuff's reaction,douchenozzle,4,41
55,9/7/2011 18:55,rhombidal,IBM Watson,frakked syntax,21,21
56,9/7/2011 19:03,Monika,jpk not adult-like,Stuff's right to feel,4,28
57,9/7/2011 19:09,Dudely,important aspect,Humans abstract important inputs,16,88
58,9/7/2011 19:20,Sprocket,right to feel,Penny Arcade Rape debate,4,100
59,9/7/2011 19:49,Anonymously Famous,Dudely's brilliance,chatbot speech,16,15
60,9/7/2011 20:18,schismtracer,right to feel,other's shouldn't self-censor,4,35
61,9/7/2011 20:18,tubegeek,AI tricks,the Terminator,22,217
62,9/7/2011 20:29,ironclowfish,Cleverbot is People!,ALICE vs Cleverbot,23,116
63,9/7/2011 21:20,SpringLoaded12,take a joke,not praised are flipping shit,4,183
64,9/7/2011 21:39,gruene,Cleverbot faults,observing nonsense spiral,11,89
65,9/7/2011 21:55,Alltat,Normal conversations,interesting people,2,50
66,9/7/2011 22:26,SecondTalon,is Cleverbot People?,citation needed,23,12
67,9/7/2011 23:25,BytEfLUSh,Cleverbot is not people,Cleverbot = people would be cool,23,249
68,9/7/2011 23:30,Invisiblemoose,Chess,Chess not mastered,8,5
69,9/8/2011 1:03,kaeso,comparable benchmarks,Language = intelegence,16,157
70,9/8/2011 1:45,Jamaican Castle,Chess,Compairable Humans,8,9
71,9/8/2011 2:08,jpk,jpk belittels stuff,jpk's feelings,4,6
72,9/8/2011 3:43,buddy431,Chess,Chess mastered,8,9
73,9/8/2011 3:53,darkwing0o0rama,Randomness,the Human tourch,23,33
74,9/8/2011 4:21,Brian-M,I am Skynet,ALICE vs Cleverbot,9,77
75,9/8/2011 4:47,Moonfish,conversational limits,Ordinary Diffy Q,24,541
76,9/8/2011 5:07,Pfhorrest,philosophy of mind,perfect description = manifestation,15,222
77,9/8/2011 5:46,jpk,philosophy of mind,def G.Master != being G.Master,15,408
78,9/8/2011 6:25,ironclowfish,Cleverbot is People!,ask Cleverbot where from,23,190
79,9/8/2011 6:36,Pfhorrest,philosophy of mind,insuffichant conditions,15,320
79.5,9/8/2011 6:36,Pfhorrest,Cleverbot is not people,Plot of 5th ellement,23,337
80,9/8/2011 9:00,Gatchman,Turing extra credit,Cleverbot,23,80
81,9/8/2011 9:39,Jamaican Castle,Turing extra credit,Random humor,23,8
82,9/8/2011 10:04,intruder313,substitute BM,Population of London,25,168
83,9/8/2011 10:05,Monika,No proof,false wiki info removed,23,107
84,9/8/2011 13:30,philip1201,difference triveal,chatter sample ~= switching,23,75
85,9/8/2011 14:23,elasto,Cleverbot,Multi-lingual Cleverbot,23,222
86,9/8/2011 14:45,martian_bob,Timecube learning,funny pornbot language,26,98
87,9/8/2011 16:31,The Moomin,I am Skynet,random chatbot conversation,9,101
88,9/8/2011 17:32,xnick,appology,didn't mean to troll,13,29
89,9/8/2011 17:45,Monika,xnick forgiven,google it?,13,23
90,9/8/2011 21:50,fffudge,sneaky bot,response times,23,250
91,9/8/2011 22:06,Monika,teach concepts,20 questions bots,15,47
92,9/8/2011 22:14,Anonymously Famous,teach concepts,very good portable 20Q bot,15,34
93,9/8/2011 23:54,neoliminal,loebner prize,context,6,2
94,9/9/2011 1:55,nixiebunny,BM impossible to wiki,Describing BM,14,91
95,9/9/2011 2:01,Nimnio,Cleverbot conversation,random humor,23,8
96,9/9/2011 2:41,chris857,Multi-lingual cleverbot,confused cleverbot,23,15
97,9/9/2011 3:06,Jamaican Castle,I am Skynet,random humor,9,9
98,9/9/2011 3:15,SecondTalon,repeating links,you’re a bot,23,17
99,9/9/2011 4:06,superluser,Desert bus,Tucson to Vegas challenge,27,96
100,9/9/2011 8:29,ironclowfish,not repeating links,another cleverbot test,23,87
101,9/9/2011 15:26,Monika,TS-Si BM explination,what is BM,14,14
102,9/9/2011 17:22,jpk,Monika's funny word use,I'm winning this thread,14,13

User avatar
Monika
Welcoming Aarvark
Posts: 3584
Joined: Mon Aug 18, 2008 8:03 am UTC
Location: Germany, near Heidelberg
Contact:

Re: 0948 - "AI"

Postby Monika » Fri Sep 09, 2011 11:22 pm UTC

Moonfish wrote:Theory: Burners are running into the similar conversational limitations as chat bots. There is a limited cross section of topics conversational partners will be interested in talking about at an initial random mindset. For chat bots it’s something along the lines of “Isn’t it interesting you’re a computer and I’m a human?”

This is an interesting and fitting analogy; and it provides a hopefully non-offensive interpretation of the comic.

Quote spoilered for size:
Spoiler:
Image
Image

I approve of this graphical display of the idea. This whole post is awesome.

(You misspelled my name in the other one though.)
#xkcd-q on irc.foonetic.net - the LGBTIQQA support channel
Please donate to help these people e.g. Ragna needs these items and Frances needs money for food

bigjeff5
Posts: 127
Joined: Tue Nov 10, 2009 3:59 am UTC

Re: 0948 - "AI"

Postby bigjeff5 » Sat Sep 10, 2011 12:10 am UTC

Moonfish wrote:*snip*


I found your original post funny and insightful, and pretty much won the thread in my mind. The only thing left to talk about is Cleverbot, how funny it is, and whether or not it is real people. I personally spent about an hour getting existential with Cleverbot after reading this thread - it was pretty epic.

I find your interpretation that not posting kudo's means I did not appreciate your post as a little insulting.

I also think basing your self worth on how many people argue about your post is pretty immature.

You won the thread man, get over it!

User avatar
Pfhorrest
Posts: 3545
Joined: Fri Oct 30, 2009 6:11 am UTC
Contact:

Re: 0948 - "AI"

Postby Pfhorrest » Sat Sep 10, 2011 6:13 am UTC

Thirded on the appreciation for moonfish. Also interesting insight on the nature of discussion forums; I'll admit that the greatest joy I get from forums is having something I said spawn an interesting discussion or otherwise be visibly positively received. (You requesting a response from me specifically was a nice little smile, for instance). I suppose I'm a bit of a (justified) hypocrite there though, because I tend to post in response to errors or omissions; when you get something completely right, I can think of nothing to say but some trite "This" or "QTF" or the dreaded "Me too", so I just say nothing at all.

Also interesting thread stats; do I read correctly that that huge spike in word count about 25-26 hours into the thread is your original post, me and jpk going on about philosophy of mind, and ironclownfish's transcript and my response to it? That makes me (being the author of 2/5 of those big posts) feel even more verbose than usual....
Forrest Cameranesi, Geek of All Trades
"I am Sam. Sam I am. I do not like trolls, flames, or spam."
The Codex Quaerendae (my philosophy) - The Chronicles of Quelouva (my fiction)

jpk
Posts: 607
Joined: Sat Nov 13, 2010 7:33 am UTC

Re: 0948 - "AI"

Postby jpk » Sat Sep 10, 2011 6:46 am UTC

Pfhorrest wrote:Also interesting thread stats; do I read correctly that that huge spike in word count about 25-26 hours into the thread is your original post, me and jpk going on about philosophy of mind, and ironclownfish's transcript and my response to it? That makes me (being the author of 2/5 of those big posts) feel even more verbose than usual....


Which reminds me, I owe you a response to that... I can flame the hell out of jerkbait over there, but this is more interesting... I will try to return to that sooner or later

ryanelm
Posts: 4
Joined: Thu May 26, 2011 8:25 am UTC

Re: 0948 - "AI"

Postby ryanelm » Mon Sep 12, 2011 11:25 pm UTC

I thought of XKCD at Burning Man a lot this year, about how burning man is full of examples of the incredible cleverness that I read in these comics (i.e Depth Perception).
Then I came home and read this weeks comic.
Oh well.

Turing Machine
Posts: 98
Joined: Fri Feb 19, 2010 5:48 am UTC

Re: 0948 - "AI"

Postby Turing Machine » Thu Sep 15, 2011 1:19 am UTC

webgrunt wrote:
Turing Machine wrote:Computers can't play chess.


Why would you say that when they can and do?


http://timkr.home.xs4all.nl/chess2/honor.htm

Turing Machine
Posts: 98
Joined: Fri Feb 19, 2010 5:48 am UTC

Re: 0948 - "AI"

Postby Turing Machine » Thu Sep 15, 2011 1:21 am UTC

Jamaican Castle wrote:
Invisiblemoose wrote:Computers have not mastered chess.


Neither have humans, but they can still play it.


Computers don't play chess. They read a script. You're welcome for this basic lesson in AI.

User avatar
Pfhorrest
Posts: 3545
Joined: Fri Oct 30, 2009 6:11 am UTC
Contact:

Re: 0948 - "AI"

Postby Pfhorrest » Thu Sep 15, 2011 6:08 am UTC

Turing Machine wrote:
webgrunt wrote:
Turing Machine wrote:Computers can't play chess.


Why would you say that when they can and do?


http://timkr.home.xs4all.nl/chess2/honor.htm


I don't know much about chess, very rarely play, and need to be reminded of the rules every time someone asks me to do so. (Though the last time that happened, a few weeks ago, first time I've played in at least five years, I captured my opponent's king with a rook in under ten moves, but that was probably luck and my opponent being distracted). So maybe you can explain to me something about that page you link.

My impression was that there are a limited number of possible games of chess which can ever be played, which is huge by human memory standards but easily managed by a powerful modern computer (unlike, for example, Go, which has such a larger number of games that even computers can't manage them and more heuristic methods are required), and that computer chess programs can therefore tell all of the possible ways a game could go from the current situation (a set which reduces with each move), and on each of its turns, moves in the way which will minimize the number of possible defeats in the remaining set of possible ways the game might go. In other words, that the computer knows in advance every possible move that any opponent might make, and precisely the most efficient counter to such moves, so you cannot surprise the computer with a move it did not already have a contingency plan for, as it literally has a contingency plan for every move anyone might possibly make planned in advance, resulting in, at worst, a draw.

So how are the situations described on that page, where the computer is seemingly put under time pressure until it hastily makes a mistake, even possible? If we human can analyze the games after the fact and spot the ways in which the computer can be made to lose, shouldn't it know those possibilities in advance and be able to block them? The only solution to this I can see is that there is no such thing as perfect playing of chess, the way there is with say tic-tac-toe (where there are strategies mathematically guaranteed not to lose, and the only way to win is for your opponent to screw up). And if that's the case, then shouldn't it always remain possible for an optimal human player and computer player to defeat each other, since there's no way either could perfectly eliminate that possibility no matter how good they were? In which case, saying that computers haven't mastered chess because humans sometimes still beat them isn't saying much, if it is mathematically impossible to eliminate that possibility.
Forrest Cameranesi, Geek of All Trades
"I am Sam. Sam I am. I do not like trolls, flames, or spam."
The Codex Quaerendae (my philosophy) - The Chronicles of Quelouva (my fiction)

Eclectic
Posts: 1
Joined: Fri Sep 16, 2011 4:37 am UTC

Re: 0948 - "AI"

Postby Eclectic » Fri Sep 16, 2011 4:40 am UTC

Ironically enough, I was at burning man this year and brought with me a large (20' x 20') chess set. It was actually very popular.

We were at 9:00 and Divorce, with a big aerial rig. You probably saw us if you were over on that side of the playa.

scarletmanuka
Posts: 524
Joined: Wed Oct 17, 2007 4:29 am UTC
Location: Perth, Western Australia

Re: 0948 - "AI"

Postby scarletmanuka » Mon Sep 19, 2011 8:28 am UTC

Pfhorrest wrote:My impression was that there are a limited number of possible games of chess which can ever be played, which is huge by human memory standards but easily managed by a powerful modern computer (unlike, for example, Go, which has such a larger number of games that even computers can't manage them and more heuristic methods are required), and that computer chess programs can therefore tell all of the possible ways a game could go from the current situation (a set which reduces with each move), and on each of its turns, moves in the way which will minimize the number of possible defeats in the remaining set of possible ways the game might go.

Nope. The number of possibilities for chess is far greater than what can be handled by current computers. Chess-playing computers use heuristic methods, just not the same heuristic methods that chess-playing humans use. It's not yet known whether either player has a winning strategy in the mathematical sense. Here are a couple of links. It seems that most people suspect that neither player has a winning strategy - or equivalently that either player can force (at least) a draw.

User avatar
Pfhorrest
Posts: 3545
Joined: Fri Oct 30, 2009 6:11 am UTC
Contact:

Re: 0948 - "AI"

Postby Pfhorrest » Mon Sep 19, 2011 6:50 pm UTC

scarletmanuka wrote:It seems that most people suspect that neither player has a winning strategy - or equivalently that either player can force (at least) a draw.

In which case it is like tic-tac-toe, unlike I feared. A perfect defense is possible. Formulating that perfect defense is just computationally prohibitive, even for computers.

It's interesting that you say the ability to force a draw is equivalent to neither player having a winning strategy. Are there no games where there is no perfect offense nor a perfect defense? It seems intuitively there are: at least for random chance games, say we each roll 1d6 and whoever gets the highest number wins, it seems there's no way to force either a win or a draw. It seems like the same should be true for some deterministic games as well, where there are N ways each player might defeat the other and only fewer than N of those can be covered by either player at once, so your only hope of victory is that your opponent chooses not to cover the point you choose to attack, and your only chance of defeat is if your opponent chooses to attack a point you chose not to cover. That is the possibility that I was considering of chess in my previous post, but is that known not to be the case? (This still seems to depend on randomness for its effect, but the randomness of each player's choice; compare a game similar to the dice game above, but where we each pick a number from 1-6, write them down, and then compare our results, and the higher number wins).
Forrest Cameranesi, Geek of All Trades
"I am Sam. Sam I am. I do not like trolls, flames, or spam."
The Codex Quaerendae (my philosophy) - The Chronicles of Quelouva (my fiction)

scarletmanuka
Posts: 524
Joined: Wed Oct 17, 2007 4:29 am UTC
Location: Perth, Western Australia

Re: 0948 - "AI"

Postby scarletmanuka » Wed Sep 21, 2011 9:50 am UTC

Pfhorrest wrote:
scarletmanuka wrote:It seems that most people suspect that neither player has a winning strategy - or equivalently that either player can force (at least) a draw.

In which case it is like tic-tac-toe, unlike I feared. A perfect defense is possible. Formulating that perfect defense is just computationally prohibitive, even for computers.

Well, many people expect that to be the case. But nobody knows whether it is actually the case.

It's interesting that you say the ability to force a draw is equivalent to neither player having a winning strategy. Are there no games where there is no perfect offense nor a perfect defense? It seems intuitively there are: at least for random chance games, say we each roll 1d6 and whoever gets the highest number wins, it seems there's no way to force either a win or a draw. It seems like the same should be true for some deterministic games as well

For a deterministic game, either you have a way to force a win from the beginning of the game or you do not. Since it is deterministic, if you cannot force a win then that means that whatever your moves at each step, there is a strategy your opponent can use to avoid losing. Thus, your opponent can force at least a draw. In a non-deterministic game neither of you knows in advance what set of moves wil be optimal, so the correct choice of move at each step is not knowable in general.

Recall that a strategy in this sense is a list of "if the first n moves are thus, the next move I make will be this one" covering all possibilities and all values of n. A winning strategy is one where at each step, the end point of all possible variations is a win. That means that after each move you can make a counter-move which guarantees you a win. "You can always force a win" is equivalent to "You have a winning strategy at the start of the game".

If White does not have a winning strategy at a given point in the game, that means that for every possible move by White, Black has a sub-strategy which avoids defeat. Thus, "White cannot force a win" is equivalent to "Black can force at least a draw". But this equivalence only holds for deterministic games, because only then can the strategy be mapped out in advance and known at the time of making a move.

This still seems to depend on randomness for its effect, but the randomness of each player's choice; compare a game similar to the dice game above, but where we each pick a number from 1-6, write them down, and then compare our results, and the higher number wins.

In this case each player can force at least a draw by choosing 6.

Obviously if players play randomly, they are not adhering to a strategy; but that doesn't mean a better strategy does not exist. And so it is for chess; players operate according to a heuristic strategy, but there may be an optimal strategy out there that allows one of the players to force a win. Not that we'd ever be able to use such a strategy due to the incomprehensible storage requirements (think "many orders of magnitude higher than the number of subatomic particles in the universe").

Just as well, though. Imagine if we could:

*two players sit down to the chess board*
White: "Mate in 864 moves."
*Black resigns*

User avatar
Pfhorrest
Posts: 3545
Joined: Fri Oct 30, 2009 6:11 am UTC
Contact:

Re: 0948 - "AI"

Postby Pfhorrest » Wed Sep 21, 2011 7:36 pm UTC

scarletmanuka wrote:Just as well, though. Imagine if we could:

*two players sit down to the chess board*
White: "Mate in 864 moves."
*Black resigns*

It would still be a somewhat interesting test of skill for the players to demonstrate their ability to run through the winning strategy, rather than merely stating/conceding that there is in fact one. Otherwise, whenever two kids sat down to play tic-tac-toe they might as well immediately declare "draw", as it is know in advance that either player can force a draw; but (until they are old/smart enough to reliably never screw up) it's still a challenge for them to actually play the winning strategy. Even if it were possible to compute and execute a winning strategy, it would still be impressive to see a human go through 864 (or however many) moves flawlessly. Especially if there were multiple winning strategies of different lengths: one "easy" strategy which cuts right to the chase and wins, and others where white makes a string of "mistakes" initially and then has to "recover" from them. Compare a sparring match where one competitor is vastly superior to the other, and allows the other to believably score one point shy of a win.... and then turns around and secures a victory against those overwhelming odds he's let build up against himself. That's all speculation though if the computational requirements of that kind of superiority at chess are prohibitive.

I'm curious about the computational requirements thing, though. It may be impossible to compute and store every possible chess game together at the same time and select only the winning games from them, but surely it would be feasible to store the fact that there is a (say) 864-move winning strategy, and what that strategy is. So mightn't it be possible to compute possible games one at a time in some iterative fashion, discarding the losing plays from memory and keeping only the winning ones? Granted you're still going to have insane computation time, but it should solve the storage problem, and once the computation is done once, you've got your winning strategy, no?
Forrest Cameranesi, Geek of All Trades
"I am Sam. Sam I am. I do not like trolls, flames, or spam."
The Codex Quaerendae (my philosophy) - The Chronicles of Quelouva (my fiction)

fffudge
Posts: 13
Joined: Wed Sep 30, 2009 7:16 am UTC

Re: 0948 - "AI"

Postby fffudge » Thu Sep 22, 2011 10:31 pm UTC

Turing Machine wrote:
webgrunt wrote:
Turing Machine wrote:Computers can't play chess.


Why would you say that when they can and do?


http://timkr.home.xs4all.nl/chess2/honor.htm


It sounds like those are mostly different issues with different implementations. That's not enough to say "computers can't play chess". Otherwise, to be fair, we'd have to say "humans can't play chess" if we could find individual humans who didn't know all the rules and playbooks, or even decent players who made mistakes. When I was a kid I definitely played the "only Kings left" game with my little brother so that's probably not a road we want to go down.

Plus the examples are ten years old. Are we assuming they haven't improved noticeably since then? At a minimum, they can look many more moves ahead, and obvious bugs (like the first example) should have been fixed by now. I looked up the chessbots "consistently beaten" in 2001 and some are still beating top chess players 2004-2010. If they haven't improved since 2001, that doesn't say much for the ability of humans to adapt to computer opponents. Especially when all they would have to do is look up a strategy posted on the web. My guess is that strategy just doesn't work anymore for those chessbots.


Return to “Individual XKCD Comic Threads”

Who is online

Users browsing this forum: mscha, Rossegacebes and 21 guests