- My Forums
- Tiger Rant
- LSU Recruiting
- SEC Rant
- Saints Talk
- Pelicans Talk
- More Sports Board
- Fantasy Sports
- Golf Board
- Soccer Board
- O-T Lounge
- Tech Board
- Home/Garden Board
- Outdoor Board
- Health/Fitness Board
- Movie/TV Board
- Book Board
- Music Board
- Political Talk
- Money Talk
- Fark Board
- Gaming Board
- Travel Board
- Food/Drink Board
- Ticket Exchange
- TD Help Board
Customize My Forums- View All Forums
- Show Left Links
- Topic Sort Options
- Trending Topics
- Recent Topics
- Active Topics
Started By
Message

Skynet is Upon Us
Posted on 1/20/18 at 10:35 pm
Posted on 1/20/18 at 10:35 pm
I don't know if this was covered since I don't check this board much, but in December Google released an academic paper on their new "Deep Mind" AI project called Alpha Zero.
An earlier version of Alpha had already beaten the world "Go" champion last year, but now they wanted to make it learn chess. If you know anything about chess, you know that computers have been beating the best players since the 90's. Garry Kasporov (top 5 player of all time) was defeated by IBM's Deep Blue supercomputer back in 1997 (which made world wide news). Since that time chess engines have gotten good enough where they can beat the best grandmasters running on nothing more than an iPhone.
Well Google didn't program Alpha to be a chess engine. With Chess engines, you typically have human input. You teach it how much to value pieces. You teach it opening theory (engines have huge opening book databases) and you give it end game databases. Chess engines are essentially "brute force" calculators that humans teach. Humans design them to value this or that piece or to calculate this or that position in some way. While they are incredibly good at chess, they have no real "intelligence."
All Google did was teach Alpha the rules of chess. They did nothing else. Google gave it no input whatsoever on chess theory.
They then said "go play" and Alpha began playing chess against itself. As it played it taught itself what was good play and what wasn't so good. They let it play itself only for 4 hours before putting it to the test.
They then took the best traditional chess engine (Stockfish) and had it play a 100 game match against Alpha. Stockfish is incredibly strong - crushing any and all humans. Keep in mind that the best human chess rating ever achieved was 2880 (obtained by Carlsen). Stockfish is rated about 3400.
Well guess what? Alpha crushed Stockfish. Alpha beat it 28 times out of 100 matches (rest were draws). Stockfish did not win a single game!
That is a stunning result. Such dominance is rarely seen. For instance, the best human (Carlsen) would not beat another top player (like Annand) 28 out of 100 with zero losses. Beating Stockfish 28 times is insane. Alpha even won a few games as black (which is much harder than winning as white).
But it's better. Alpha completely bamboozled Stockfish to the point it was funny (and brilliant). He even got Stockfish into a Zugzwang (it's a chess position where the opponent is totally boxed in and any possible moves are losing moves).
If you want to watch that "Zugzwang" match, this Russian guy on Youtube replayed it and gave a good analysis of it: LINK
An earlier version of Alpha had already beaten the world "Go" champion last year, but now they wanted to make it learn chess. If you know anything about chess, you know that computers have been beating the best players since the 90's. Garry Kasporov (top 5 player of all time) was defeated by IBM's Deep Blue supercomputer back in 1997 (which made world wide news). Since that time chess engines have gotten good enough where they can beat the best grandmasters running on nothing more than an iPhone.
Well Google didn't program Alpha to be a chess engine. With Chess engines, you typically have human input. You teach it how much to value pieces. You teach it opening theory (engines have huge opening book databases) and you give it end game databases. Chess engines are essentially "brute force" calculators that humans teach. Humans design them to value this or that piece or to calculate this or that position in some way. While they are incredibly good at chess, they have no real "intelligence."
All Google did was teach Alpha the rules of chess. They did nothing else. Google gave it no input whatsoever on chess theory.
They then said "go play" and Alpha began playing chess against itself. As it played it taught itself what was good play and what wasn't so good. They let it play itself only for 4 hours before putting it to the test.
They then took the best traditional chess engine (Stockfish) and had it play a 100 game match against Alpha. Stockfish is incredibly strong - crushing any and all humans. Keep in mind that the best human chess rating ever achieved was 2880 (obtained by Carlsen). Stockfish is rated about 3400.
Well guess what? Alpha crushed Stockfish. Alpha beat it 28 times out of 100 matches (rest were draws). Stockfish did not win a single game!
That is a stunning result. Such dominance is rarely seen. For instance, the best human (Carlsen) would not beat another top player (like Annand) 28 out of 100 with zero losses. Beating Stockfish 28 times is insane. Alpha even won a few games as black (which is much harder than winning as white).
But it's better. Alpha completely bamboozled Stockfish to the point it was funny (and brilliant). He even got Stockfish into a Zugzwang (it's a chess position where the opponent is totally boxed in and any possible moves are losing moves).
If you want to watch that "Zugzwang" match, this Russian guy on Youtube replayed it and gave a good analysis of it: LINK
Posted on 1/20/18 at 10:38 pm to AUstar
We're already fricked.
Might as well shut up and hope the T-1000s use lube
Might as well shut up and hope the T-1000s use lube
Posted on 1/20/18 at 10:38 pm to AUstar
I for one welcome our new robot overlords.
Posted on 1/20/18 at 10:39 pm to AUstar
Skynet vs. truck nuts at Sonic
Tonight
Tonight
Posted on 1/20/18 at 10:43 pm to AUstar
All your base are belong to us
Posted on 1/20/18 at 11:03 pm to AUstar
I would be interested to know when the winning games occurred. I would imagine alot of them happened after the 50 game mark, as Alphas learning curve flattened.
I would also be interested to know how many games Alpha was able to simulate in the 4 hour period. Did they track the "real time" of those games? Take the number of games simulated x the "real time" of those games and get X. Then compare that to Malcolm Gladwell's Outliers theory that it takes 10,000 hours to become a master at something.
I have a couple books on AI I want to tackle this year, just haven't started on them yet. I read The Soul of a New Machine in Nov. as a primer to read some AI theory.
The future of AI is really interesting, and I am looking forward to seeing what happens when they start trying to tackle the real world problems, like hunger/disease etc. using these AI computers.
Like anything though, that type of power in the wrong hands could be catastrophic. The future is a bright scary place to look forward to.
I would also be interested to know how many games Alpha was able to simulate in the 4 hour period. Did they track the "real time" of those games? Take the number of games simulated x the "real time" of those games and get X. Then compare that to Malcolm Gladwell's Outliers theory that it takes 10,000 hours to become a master at something.
I have a couple books on AI I want to tackle this year, just haven't started on them yet. I read The Soul of a New Machine in Nov. as a primer to read some AI theory.
The future of AI is really interesting, and I am looking forward to seeing what happens when they start trying to tackle the real world problems, like hunger/disease etc. using these AI computers.
Like anything though, that type of power in the wrong hands could be catastrophic. The future is a bright scary place to look forward to.
Posted on 1/20/18 at 11:05 pm to BRgetthenet
In a full on brawl I’ll take some Ascension parish baws anyday over robots.
Posted on 1/20/18 at 11:24 pm to JackieTreehorn
Someone explain to me why it's harder to win with black vs. white. Does white always have first move?
Posted on 1/20/18 at 11:29 pm to TbirdSpur2010
quote:
We're already fricked. Might as well shut up and hope the T-1000s use lube
Don't worry about T1000s showing up from the future, the Boston Dynamics Atlas project and others like it are already here and pretty damn scary.
LINK
Posted on 1/20/18 at 11:33 pm to AUstar
We need to shut down these mad scientists creating this AI before it’s too late
Posted on 1/20/18 at 11:35 pm to AUstar
Is this the same one that developed its own language?
Posted on 1/20/18 at 11:37 pm to rmnldr
quote:
scientists
Scientists are all alike.
They say they’re working for us. When, what they really want is TO RULE THE WORLD!!!!
Posted on 1/20/18 at 11:38 pm to WaWaWeeWa
shite, I can beat Alpha right now. Just before I sit down to play I will tell it to play tic tac toe against itself. Checkmate.
Posted on 1/20/18 at 11:42 pm to EA6B
quote:
Don't worry about T1000s showing up from the future, the Boston Dynamics Atlas project and others like it are already here and pretty damn scary.
That vid hits on two things that so many people don't seem to understand.
Manufacturing jobs are NEVER coming back b/c robots are going to be doing the work and it doesn't matter how many guns your redneck cousin has stored up when they send a tiny drone with a bomb attached to blow his arse off the planet.
Even scarier, that's the stuff they're willing to show us.
This post was edited on 1/20/18 at 11:43 pm
Posted on 1/20/18 at 11:44 pm to BRgetthenet
(no message)
This post was edited on 6/6/20 at 12:34 pm
Posted on 1/21/18 at 12:23 am to Ham Solo
@Ham
Yes white always has the first move.
@WaWaWeeWa
Negative
Yes white always has the first move.
@WaWaWeeWa
Negative
Posted on 1/21/18 at 12:38 am to Ham Solo
quote:
Someone explain to me why it's harder to win with black vs. white. Does white always have first move?
Because whitey is always putting them down.
Posted on 1/21/18 at 12:49 am to WaWaWeeWa
quote:
Is this the same one that developed its own language?
None ever really developed their own language. That was just some overly excited journalism.
Popular
Back to top

14







