Deep Fehler

rooting for you to win the challenge christian :)
then, by the transitive property, I will have won a game against all the bots (I think I won a game against you at some point).

Christian: I missed your earlier comment (i was away from LG for a few days).

I agree that there needs to be a broad base of games to explore and learn from, with conclusions being drawn bottom-up, rather than just pronounced by humans with pretensions (on either side).

I'm interested in games that are large, but that can be solved completely. It is fascinating to see the difference between an *extremely* strong algorithm, and a perfect oracle.

I did that experiment with Roel van der Goot's Awari program Softwari (which is probably many many classes stronger than any human who ever played the game over the past 4000 years) against Jon Romein's full enumeration Awari database. It was delightfully sickening. I wasn't sure how often Softwari could hold the game-theoretic draw, but it quickly became apparent that it would probably lose a 100 game match 0-100, without any games even being close. The grand over-champion Softwari was losing a couple of pebbles (compared to optimal) every 2 or 3 moves... The difference between perfection and very very very good is simply *enormous*.

There was a time when i doubted that a 3000 chess rating would ever be achieved, due to the plateauing skill curve. But now, with a better understanding of perfect oracles, i wouldn't bet against a 10000 rating being conceivable.

The trials of mere humans are relatively uninteresting to me in comparison. ;-)

Perhaps we really should discuss it over a beer some time. It might be a fun conversation that could go on for hours...

Coffee for me please :)

I'm interested in the trials of mere humans because I invented games for humans. I'm interested in bot preformance mainly within that context. The CodeCup Challenge will be won by a program, but with only other programs to compare it with, this predictable outcome will imo. lack significance. For one it would hold for any game and thus shed no light on why some games are more difficult to program than others.

The ultimate killer for any game is solving it, though players still seem to enjoy checkers or awari. But the enormous difference between a very strong algorithm and a complete solution seems the same as between 'very large' and infinite - there's just no comparing the two. The solution can't go wrong and strong algorithms face the same defeat against it as weak ones.

Actually I'm interested in games that are small but cannot as yet be solved, nor effectively 'captured' by an algorithm.

Are humans fighting a losing battle? I don't know. Extrapolating on advances in calculation speed and clever approaches (where solving is not an option) I'd say they well might. But humans may still be able to keep an edge in certain games. If computers beat humans at chess, but not in the same measure in Havannah, there must be reasons for that. One is of course the sheer amount of effort put into chess programming, but at the same time the human effort put into chess is also much larger.

Any news on the challenge? Do we have a date yet? (Sorry, I know this question keeps being asked, but a lot of us would really like to follow the games)

I'm still training and trying to get more grip on strategy and I'm looking forward to an interesting match. Tony would be the right person to ask about a date. In fact I'm rather curious too :)

Well, Tony, what do you say?
By the way, has it been oficially decided which bots are competing?

I've mailed Ton on May 18th, asking (again) for an account number to deposit the 1000 euro, and an outline of the state of affairs concering the organization. I'd like a date too, to see if I can generate some additional interest in the event.

Ok, looks like it's going to be end of August, beginning of September, at iGGC, courtesy of Arty Sandler. I count Castro, Wanderer and Lajkonik in, but I haven't heard anything yet about the German entry. Maybe Ingo can shed some of the light he so abundantly radiates, on this subject ;-) .

Come to think of it, if it plays like a bot, and base-10 at that, good chance it's a bot ... and 'kenzopower' (no info) plays like a bot. Kenzo is a dice game ... hmmm, Monte Carlo. Is it made in Germany I wonder?

Actually 'no info' isn't quite true: it plays only Havannah. Ganz klar :)

_Richard Lorenz wrote:
Also, depending on how many games I will actually play against Freeling, there are some thoughts of doing some anti-human tuning in an attempt to steal a win.

Ingo wrote:
You are not alone with this attitude._

Ah, yes, some 'anti-human tuning'. So why did nobody think of that before, lol. If there's to be a secret weapon, at least make sure it exists ;-) .

So, does anyone have any idea who's behind this failed Turing test?

I send a message to Kenzopower, and it turns out I was wrong:

"Hi, I’m not a bot :-) I disovered Havannah on the 8-size Ravensburger edition and love it. I still have some trouble to adjust to the long delays between moves here on little golem and have not player on 10-size boards before. I hope to get beyond a bot-like play. Regards, Max"

To which I replied:

"How about that. Hope I didn't offend you, you're not that bad, but your play so far appears to revolve around tactics, and that's precisely where the current bots are at their best :) . christian"

Being called a Failed Turing Test has got to sting. An ad machinam attack?

You might consider what I actually said - I didn't call anyone a failed Turing test, but only figured I played against a bot. Figuring out the difference between a human and a bot is 'a failed Turing test'.

I was wrong in my assumption. I apologized to Max. Any more insinuations or character smears?

Regarding the German bots, this is the latest I heard from Johannes Waldmann:

_At this moment I am not sure about the German Bots.

I forwarded your announcement to Deep Fork.

Ring-von-Fehler: I had a student work on some improvements (doing playouts on a CUDA programmed GPU) but so far I didn't have time to evaluate them properly, and merge them back with the main program. I may do this in during the summer - but I also hear that this is the time where people usually go on holidays ..._

Deep Fork is a program by Thomas Reinhardt.

Ingo ("the bots will pull off your undertrousers early enough") Althofer informed me that he will not participate in any of the teams for the challenge and will follow the event as an interested spectator.

I hope both bots will participate of course, if only to enhance my chances ;-) .

Quotes from Ingo, who leaves the programming to others:

"You made your offer (in 2002) that no bot will beat you in a single game (out of ten games) within ten years. No programmer – at any time – made a public announcement that (s)he will give you sparring time before such a match."
There were hardly any answers the first five years. You seem to demand specific conditions, in retrospect, for a match for which there were no opponents yet, just silence and ignorance.

"Sometimes I have a memory like an elephant. Between 2002 and 2009 Christian was in big mouth mode, concerning computer chances in Havannah. Read the original announcement from 2002, or his early postings in 2009 here in the forum. Then, gradually he switched into more polite gears."

Programmers only dared to take up the challenge after MCST became 'promising'. Before that they were in silent mode. I must have seemed 'loudmouthed' by sheer contrast.

I got back to 2009 and found a thread titled Christian Freeling and the Bots. If that one is representative, please show me where I was less than polite.

Now let's get over this "I don't answer because you're not polite" shall we. You can declaim all you want, but you cannot explain why the German bots are silent (which may seem to make you loudmouthed by contrast). It seems less than polite and less than courageous.

*Das Schweigen der -Lämmer- Roboter*

Although Ingo Althofer was never slow to answer any of my entries in the Word Association Thread, that may change. His latest entries amount to _"! (no more words)"_ and _"writing is silver, silence is gold"_. We can but wait.

One and a half years ago I posted a question. This question:

_"Can someone please shed some light on why there are currently no real bots playing Havannah? I can't say that I'm bothered by Ring of Fire, because a game against it takes as long as it takes to form a 7-stone ringframe. But what's the point of entering a bot with a 1291 rating and a known bug, I wonder?
And what's the point of withdrawing all the real bots?"_

Ingo Althofer replied:

"_Christian, *you should not make such a comment*. Science sometimes is going strange ways, and people from outside (like you) do not understand the motivation.

Simply wait until late 2012: the bots will pull off your undertrousers early enough ;-) _"

The patronizing and elitist attitude obviously sought to comply with the 'strange ways' of science as understood by insiders (or at least according to Althofer), and not by _people like me_.

And what, I wonder is so mysterious about the motivation: obviously the German bots aren't up to it and you're afraid to lose face. To people who don't understand the mysterious ways of science, like me, your 'golden silence' simply looks like a less than courageous retreat into an ivory tower.

At least I got the courage and _decency_ to put my money where my mouth is.

@christian
It will be very interesting to follow this challenge. Keep us posted when the date and circumstances are known to you. I assume it is still unclear which bots you will face?

Good luck! Or better, since luck is not involved, Good intelligence!

Castro, Wanderer and Lajkonik for sure. Regarding Ring von Fehler Johannes Waldmann wrote:

"Ring-von-Fehler: I had a student work on some improvements (doing playouts on a CUDA-programmed GPU) but so far I didn't have time to evaluate them properly, and merge them back with the main program. I may do this in during the summer - but I also hear that this is the time where people usually go on holidays... "

and regarding Deep Fork:

"I forwarded your announcement to Deep Fork.", which, in person, is Thomas Reinhardt.

That was two weeks ago - nothing from Thomas yet.

July 3 the organizing committe will have a meeting to discuss several aspects of the match, including a closing date for entries.

_"Simply wait until late 2012: the bots will pull off your undertrousers early enough ;-)"

Ingo._

There'll be an update shortly, but guess what: the german bots bolted out, their undertrousers full of eh ... Ingo's announcement, lol.

It seems like there should be a maximum on how many times a quote can be repeated in a single thread by the same person.

Alternatively you could limit the number of uses of the word 'undertrousers'. I do not like that word.

Not a very pretty word is it. Like recidivism. "No sir, that's one bonehead word, but that ain't me any more."

Ingo stated I was 'loudmouthed' about the challenge in the time that no-one picked it up (because well, you see, they couldn't - there was no MCTS yet). It's the usual trick: distract from the content by questioning the TONE - I've encountered this cowardice all my life.

Then MCTS came and suddenly the AI community appeared to have a chance. Where I was totally neglected before, now there was a sudden interest. That was good of course, only ... not that good, and Ingo's announcement was unnecessarily rude, and very inappropriate, because he speaks for others instead of taking up the challenge himself. Now these others don't participate either. Is it the TONE, maybe? Or just cowardice? 'Feigheit' in German.

AI students profit of my game by writing interesting articles and MSc theses and the like, supporting their careers. What do I get?

I'll tell you: nothing but the recognition (by some). If I lose one game, I pay 1000 euros - the money has already been deposited (to inform those who like to question such things). If I win all ten games I get nothing.

And because of that I should of course shut up.

Undertrousers, undertrousers, undertrousers.

@christian:

> ""

You get infinite glory for having invented/discovered the game. The more the interesting articles and the MSc and PhD theses, the more you gain.

(correction, that should be:)

> "AI students profit of my game by writing interesting articles and MSc theses and the like, supporting their careers. What do I get?"

You get ...

I agree completely, and I'm happy for them and for me. But being presented with substitute opinions about my chances by someone who doesn't care to take up the challenge, and the accompagnying mantra about the flight 'artificial intelligence' is going to take if I just wait ... , is annoying. A faster clock speed doesn't turn impotence into 'intelligence', and I weep for a humanity that trusts 'intelligent programs' to take market decisions that may affect millions, if the same intelligence should turn out to be incapable of winning a pencil and paper game.

You sound like a less profane version of Mark Steere, down to the repeated postings into empty air, while everyone with any sense sits around being uncomfortable with the spectacle.

You made a wonderful game, Ingo made a snide comment about your trousers that he shouldn't have, and you've certainly responded more than enough. Leave it at that. This isn't rec.games.abstract.

You're right, call me when putting your money where your mouth is becomes a virtue again ;-)

Those Germans are clearly horribly inefficient - I can optimise the taunting by a massive 23%!

Underpants, underpants, underpants!

Up to white_18 this was actually an interesting game.

So not all losses are equal? Strange that it should put up a fight only after throwing the game away. I expected random play once it was all equally futile.

If the man vs. machine match is you vs. Fehler, my 10e is on the human!

I'm not sure whether will be RvF participating, nor whether this is the strongest implementation. I know the bots are weak at ring recognition, but I'm certainly not going to be 'playing for that'. This ring threat was a mere tempo/shape move and I was very surprised it missed it. Castro, Lajkonik and Wanderer probably wouldn't have, though they tend to miss the somewhat deeper hidden ones too. At least the versions I played against. Achieved results are no guarantee :)

In this game Lajlonik shows its ringhappy approach and indeed, at black_91 it replies correctly to the same threat that RvF missed. The fact that it is a double ring threat is unfortunate, but obviously Lajkonik has the better spectacles.

So, Christian, have you as yet ever lost a game to one of the bots? If so, have you ever lost one where you were paying attention?

Yes and obviously not ;-)

How well do you think the bots will play in the faster time control? On here, they take several minutes per move.

If the evaluation is not correct, clockspeed and time may not matter all that much. 'Intelligence' implies an interest, a plan, a strategy to execute it. Bots don't have an interest, don't have a plan, and execute an algorithm. That may be a smart algorithm, but is it intelligence?

In a wider context, I distrust 'intelligent systems' and a world wherein 'human' is primarily associated with 'error'. First beat me at a pencil & paper game ;-)

I run single-threaded on LG but can use multiple threads in tournament conditions. This means a similar amount of work done in shorter time per move.