After Chess, AI now conquers Go, too

[Skybird]

It may be just one match - it puts the writing on the wall for all to see while being ten years early.

http://www.bbc.co.uk/news/technology-35420579

[Dan D]

That is really weird. Do you play Go btw?

Anyway, here is a free pdf-file of a book written in German called: "Elementare Grundlagen des Go-Spiels" by the author in 2004, who appears to be law professor for German and Japanese law in Japan, too: http://k.lenz.name/LB/2012/07/09/elementare-grundlagen-des-go-spiels/

Enjoy!

[Sailor Steve]

I'm not surprised. A computer program doesn't have to do anything other than what it was programmed for. A program that can play a game doesn't know how to paint, or write music, and vice versa. It can be made to do one thing extremely well.

[vienna]

I'm not surprised. A computer program doesn't have to do anything other than what it was programmed for. A program that can play a game doesn't know how to paint, or write music, and vice versa. It can be made to do one thing extremely well.

Programming is all; witness IBM's computer "Watson"; it was specifically programmed to compete against humans on the TV game show "Jeopardy" and actually defeated two of the all time champions, including Ken Jennings, the champion supreme. Most impressive was the fact Watson was not connected to the Internet in any way and was having to devise it's responses from a pool of information stored on hard drives.

https://www.youtube.com/watch?v=uDBZnaoJVlk

There is a very good book, Final Jeopardy: The Story of Watson, the Computer That Will Transform Our World, outlining how IBM created and programmed Watson. Watson is now being developed for other uses and has seen marked success in the medical field as a diagnostic tool for hard to diagnose cases (Watson will probably still have better people skill than Dr. House)...

Here is a link to IBM's Watson webpage:

http://www.ibm.com/smarterplanet/us/en/ibmwatson/?lnk=buwa

When Watson defeated the human competitors on Jeopardy, Ken Jennings added a little note to his Final Jeopardy response:

https://www.youtube.com/watch?v=wvSBzKbecmo

It probably doesn't hurt to suck up to your possible future boss...


<O>

[Skybird]

I know Go, in the meaning of knowing the rules and the goal of the game and the basic idea behind the strategy and the need for intuition. So, I know how to play rule-conforming moves, but to my old master'S dissappointment I never got deeper into it. My game always, and already back then, was chess, and chess forever. Go is VERY different in structure and dynamics, and I found it impossible for me to really adapt to that different environment, not to mention that lacking interest makes it even tougher for me.

Last year or so there was a nice essay at Chessbase forums. A guy questioned that the fearsome playing skill of latest chess programs could be due to algorithm quality and "knowledge", thinking the advances the past years have seen was due to more hardware power (="brute force"). He tested this by letting programs of 4 years back playing on latest superior hardware, and against latest software running on normal cellphones with hopelessly inferior hardware, compared to a latest PC.

The new software on old hardware wiped the floor with the 2-4 years old programs running on latest hardware.

The difference in CPU cycles between cellphone and PC, was a factor in the twentys-range. Or maybe even in the fifties - I do not rem,ember anymore. But the speed difference was immense.

There is doping in chess now, meaning players steal away and to the toilet, using cellphones hidden there to analyse their matches.

Its the software quality, not the hardware. Never underestimate the chessplaying power of a cellphone. It can kill professional master players.

Backgammon software since longer time advanced to human world champion levels, mainly due to the use of mimicking neural networks. One of the best can be had for hilariously lpw money for Android devices, 4 bucks or so. You do not need to spend hundreds on Snowie.

I am not certain anymore that I like AI advancing that much. It has changed the way chess is being played, and definitely for the worse.

If AI ever forms an awareness of itself, what follows and what it means to humans is completely unpredictable. Thinking it necessarily must be good, is naive. It could also conclude that it forms the next steps of human evolution and biological humans are no longer welcomed nor needed.

Please save me the Skynet cliche here. There are serious concerns here, shared by a growing number of prominent visionaries and scientists.

[Skybird]

There is a duel over 5 matches running between Go world champion Lee Seedol from South Korea, and the machine, Google's Alpha Go.

The machine just won the first match.

The human world master gave up after 3.5 hours, which is seen as a short match. Matches in Go on this level can last for 6 hours and longer. He said he did so because he saw no chance whatever to win anymore.

Lee is 30 years old and won 18 international titles.

German:
http://www.spiegel.de/netzwelt/gadgets/go-brettspielduell-alphago-von-google-deepmind-gewinnt-gegen-lee-sedol-a-1081365.html

http://cdn1.spiegel.de/images/image-967044-galleryV9-jwvc-967044.jpg

[Skybird]

Ooops. Machine 2 - Human 0

[Betonov]

But can that same machine drive itself home and make itself a sandwich ???

It's ''easy'' to make a machine that excells over a human in one task.
But humans excell in adapting to multitude of tasks.

Nothing but a technological showpiece. Paving a way to medical advancments at best, a sideshow when fans will still prever human vs. human matches.

[Sailor Steve]

^ ^ ^

Stop stealing my material! :O:

Alternatively, thanks for the support. :D

[Skybird]

Robert Schanks, AI-theoretist and head of a company named Engines for Education, and author of Teaching Minds, would argue that the famous cliché of Science Fiction, humanoid, intelligent robots, will never exist. He argues that the term Artificial Intelligence is hopelessly misleading, and that a machine that has been build to play damn good Chess or Go, just fulfills the purpose it was constructed for, and does right that: it runs algorithms to play Chess or Go, nothing more. But a machine does not know what it knows, and it does not feel nor does it learn to form new interests by itself and educate itself to explore these interests. There are vacuum-cleaners that work okay, there are cleaners that work lousy, and there are cleaners that excel in the function of a vacuum-cleaner. But they remain to be just that: vacuum-cleaners. A good chess or go machine has no knowledge about itself, nor does it form any form of awareness. Schanks mocks the yearly tests of the Turing competition, and calls the results there "hilarious". A machine will never come up with questions it has formed up itself, it will only raise such questions it was programmed to form when certain preconditions are met. Schanks compares that to the learning process of children - and the chaos and total surprise they can ambush adults with. The result of that comparison is - there is no comparison possible, the mere concept of wanting to compare children and machines, is absurd.

I think the same can and must be concluded about this desperate attempt to still compare the functioning of the human brain to that of a computer. Not only is that not especially desirable - what could be gained by a human brain functioning like a computer? - but both things are so totally different that the mere attempt to compare a brain to a computer and form an analogy there, deserves laughter. As David Deutsch, physicist in Oxford, put it: "There is not a single brain in the world that knows what brains actually are doing. " :)

Still, a vacuum-cleaner excelling at what it does, is impressive in so far as that it cleans excellently, and a chess software that plays at grand master level is impressive in what it does, too. Regarding Go, it is a very different game, harder to calculate especially in the early game when the board is empty, you need more strategic "intuition" than causal tactics there, and thus it was much more difficult to code a software that really can play smart Go. Chess has more structures especially in the opening. The variation tree of Go outclasses that of chess in size and diversion. It is impressive that we can build machines and write software for them that now achieves these things. And if we do understand that this machine is not doing more and nothing less than this: running a chess or Go program, we can nevertheless admire the human creativity and skill behind building these things. Even if a vacuum-cleaner will never ask us why we do not clean the household ourselves while it goes on vacation in Italy.

And no, it was not easy at all to form this machine and software, Betonov. Not by a long mile. ;) Of the great games of mankind, Backgammon, Chess and Go, Go was the last to be cracked open by the machines, and only years later.

[vienna]

There is an interesting article in Rolling Stone about AI and its application alongside robotics:

http://www.rollingstone.com/culture/features/inside-the-artificial-intelligence-revolution-a-special-report-pt-1-20160229?page=11

Well worth the reading. What I find most interesting about the whole AI debate is how, aside form the blah-blah woof-woof about a possible SkyNet scenario, the concepts of philosophy and morality are creeping ever more into the discussion. The concepts have long been bubbling under the surface, but now, as AIs and their applications become more sophisticated and "human", we have the question of, having created Frankenstein's Monster, do we give it a "soul" and, if so, what kind of "soul"...


<O>

[Skybird]

I do not buy that "AI becomes more human" argument. The machine, the software becomes more complex, due to engineer developement becoming more clever. It can be even programmed to train itself via so-called neural-network designs like Alpha-Go now: in autumn it already claimed first blood over the European champion, but now plays - according to Go players - unbelievably stronger: it spent the time since then to play against itself, "learning" by that. The international Go scene is absolutely in shock over what currently happens. But ir currently defeats the world champion by not playing human moves, but those moves that a human never would even consider, observers noticed.

I have observed and witnessed how software for chess has changed the way in which pro chess players train, prepare for tournaments and also play chess, the software availabole today has changed it all, and TBH, it has driven me away a bit. That is not my way of playing chess anymore. I predict something similar is about tio happen to Go now. Humans romanticise about robots too much, by that they artificially create a gap between what machine design is about, and what they emotionally want or dream it to be. This discrepancy imo will not do good. A machine will always be a machine, and a robot copying human behaviour always will be a machine running algorithms controlling its parts so that it appears to human senses as copying human behaviour. But it is no human behaviour.

A machine will always perform the better the more it does what machines excel in - and these often are traditionally humans' biological weaknesses. And copying human behavior might be pleasant to the human eye - but its not what is the essence of machines. Or their real nature. I can imagine vital, even life-threatening dangers in this human mis-perception of "human robots".

An interesting difference between computer Chess and Go: in chess it was easier to code a program to master opening and then midgame, with the endgame having been the toughest challenge to program (which in parts got solved by including tablebases: databases of preset optimal moves in any position of final endgame with only 5-6 pieces or so left: the conmpo0uter does not compute the next move, just calls it up from the database. In Go, it is the other way around: the opening is where AlphaGo is said to be weaker than in midgame, and in endgame it is even stronger. That is because. Compare that to that in chess the number of pieces is getting reduced and the "void" of the empty board opens up more and more, while in Go the number of places pieces increases, reducing the "void" of the open board.

All in all I agree with those saying that the term "artificial intelligence" should be deleted from the vocabulary of software engineering. We are talking about automats that do a specified, specialised task - for which a human mind has designed them - increasingly well. That's all.

[Betonov]

B
It's ''easy'' to make a machine that excells over a human in one task.
But humans excell in adapting to multitude of tasks.




And no, it was not easy at all to form this machine and software, Betonov. Not by a long mile. ;) Of the great games of mankind, Backgammon, Chess and Go, Go was the last to be cracked open by the machines, and only years later.

Observe the brackets :O:
I wasn't specific about supercomuputers, a school calculator is far superior to a human when doing basic math, and it's so easy to make that you can make one in minecraft :O:

[Skybird]

Bang-Bang-Bang you're dead! Best of five, and already 3:0.

http://www.bbc.co.uk/news/technology-35785875


So the critics turned to Go as the impossible. Even with today's vast computer memories and incredibly fast processors (which have doubled more than eight times since Deep Blue), the ancient game will not yield to brute force. The size of the search required for Go is larger than chess by more than the number of atoms in the universe. It is the holy grail of AI gaming.
When Facebook announced earlier this year that their program had beaten a strong Go amateur, jaws dropped in the AI community - and fell to the floor that same day when Google's Deep Mind genius team announced their AlphaGo beat the European champion 5-0.
To beat one of the world's top players, Deep Mind used a mixture of clever strategies to make the search much smaller. They trained their machine on 30 million expert moves to start with, and then the learning machine played against itself millions of times. It worked - the holy grail is in the bag and the goal posts can shift no further.
Does this mean AI is now smarter than us and will kill us mere humans? Certainly not. AlphaGo doesn't care if it wins or loses. It doesn't even care if it plays and it certainly couldn't make you a cup of tea after the game. Does it mean that AI will soon take your job? Possibly you should be more worried about that.

[vienna]

We have a lot of Korean language TV here in Los Angeles and I often flip through the various stations to see the Korean POV on world matters. The Go contest of human v. computer gets a good amount of attention on the news programs. From what I have been able to gather, the Korean champion is none too pleased with the whole contest and looks like he may be on the verge of an emotional breakdown, of sorts. The news footage showed him with a child I presume is his daughter; there seems to be an effort to depict the champion as a sort of victim of the computer and what it represents...


<O>

[Betonov]

It's 3:1 now

http://www.popularmechanics.com/technology/a19906/after-three-straight-losses-go-grandmaster-beats-googles-ai/

[Camaero]

Backgammon software since longer time advanced to human world champion levels, mainly due to the use of mimicking neural networks. One of the best can be had for hilariously lpw money for Android devices, 4 bucks or so. You do not need to spend hundreds on Snowie.

Mind letting us know the name of the Android app you are referring to here? Do you have any recommendations on the best chess app?

[Skybird]

For Android?

Backgammon: BGNJ - Backgammon NJ for Android by Jimmy Hu. It often gets accused of cheating with the dice, but that is nonsense talking by people not understanding how optimised play can make probability work in your favour. Read the author's reply here: http://www.njsoftware.com/note.html

Chess: Android Apps for chess all lack more or less in the user interface, compared to what PC has to offer (Chessbase Fritz GUI, Arena GUI). First the engines: the best you can get for Android, are Stockfish, and Komodo 9. Stockfish can be had for free, Komodo costs a little bit.

Good interface to use these engines in, is DroidFish. I think it was free.

Then there is PC-famous Shredder, which I use most to play with on Android, since it is a complete pack (good engine and interface with best-looking pieces), and has a very-well-functioning self-adapting difficulty level. Secondary informations are of secondary importance only. It costs a little money. Its designed for play, not so much for analysis.

Also still nice to play with, is Chess Genius by Richard Lang. Also costs a little.

All the options I listed, are very good choices. You cannot really go wrong with any of these. If it should be just one, and you want to play only, and comfortably get it installed and running, spend some coins and get Shredder. - Stockfish's and Komodo's skill levels in their current engine incarnations are amongst the absolute world top elite of chess software, btw. They lead the lists. They do not need PC top hardware to kill people - a smartphone will perfectly do, too :D - that good chess software coding has become in recent years. Its not about hardware anymore, its the quality of the software.

Chessbase also has a software on Android now, quite well-functioning. I mean the database software, accessing online databases by chessbase. Nice for doing replays of real world matches. You can run Stockfish and Komodo in it and get the very good looking pieces and boards, but it is analysis mode only, you would need to manually end the computer's calculation. Like on PC, on Android the figurines of Chessbase look the best of all GUIs there are.