Forums

Alphazero vs Leela zero: when will this duel happen?

Sort:
barretoff

After several months of development and based on the games made here at chess.com, it has been proven that Leela has already overcome Stockfish in strength. In other words, in theory, Leela is already ready to face her "father" or "older brother".

That said, when will the duel between Alphazero and Lc0 be? Is there a possibility of this happening? What do the teams of developers for the two programs think about this? Does anyone have this information?

IMKeto

No idea and don't care.  Engine chess bores me to tears.

barretoff
IMBacon escreveu:

No idea and don't care.  Engine chess bores me to tears.

I don't think like you do. The neural networks are very interesting, the movements they make during games are often surprising, I think it's really cool.

Martin_Stahl
barretoff wrote:

After several months of development and based on the games made here at chess.com, it has been proven that Leela has already overcome Stockfish in strength. In other words, in theory, Leela is already ready to face her "father" or "older brother".

That said, when will the duel between Alphazero and Lc0 be? Is there a possibility of this happening? What do the teams of developers for the two programs think about this? Does anyone have this information?

 

Pretty sure the AlphaZero team (DeepMind) have moved on to other areas of research using the technology. Can't say they might not revisit in the future, but I don't imagine they are even thinking about it at all.

barretoff
Martin_Stahl escreveu:
barretoff wrote:

After several months of development and based on the games made here at chess.com, it has been proven that Leela has already overcome Stockfish in strength. In other words, in theory, Leela is already ready to face her "father" or "older brother".

That said, when will the duel between Alphazero and Lc0 be? Is there a possibility of this happening? What do the teams of developers for the two programs think about this? Does anyone have this information?

 

Pretty sure the AlphaZero team (DeepMind) have moved on to other areas of research using the technology. Can't say they might not revisit in the future, but I don't imagine they are even thinking about it at all.

In fact, I've been reading around that Deep Mind is applying the concepts of Alphazero in medicine. Maybe they are no longer interested in chess, but it costs nothing to dream and it would be very interesting the duel between A0 and the program created based on it, Lc0.

Marie-AnneLiz
barretoff a écrit :

After several months of development and based on the games made here at chess.com, it has been proven that Leela has already overcome Stockfish in strength. In other words, in theory, Leela is already ready to face her "father" or "older brother".

That said, when will the duel between Alphazero and Lc0 be? Is there a possibility of this happening? What do the teams of developers for the two programs think about this? Does anyone have this information?

No need to do that match because daddy is retired now.

IMKeto
barretoff wrote:
IMBacon escreveu:

No idea and don't care.  Engine chess bores me to tears.

I don't think like you do. The neural networks are very interesting, the movements they make during games are often surprising, I think it's really cool.

I think the technology is awesome.  And the fact that someone got a learning machine to go from beginner to beating the best chess engines in 4 hours is impressive to say the least. 

But the reason why watching them play doesn't interest me?  I dont understand the moves.  I have no idea why 2700-2800 GM's make the moves they do.  So I am definitely wont understand what something rated 3500 is doing.

Marie-AnneLiz
IMBacon a écrit :
barretoff wrote:
IMBacon escreveu:

No idea and don't care.  Engine chess bores me to tears.

I don't think like you do. The neural networks are very interesting, the movements they make during games are often surprising, I think it's really cool.

I think the technology is awesome.  And the fact that someone got a learning machine to go from beginner to beating the best chess engines in 4 hours is impressive to say the least. 

But the reason why watching them play doesn't interest me?  I dont understand the moves.  I have no idea why 2700-2800 GM's make the moves they do.  So I am definitely wont understand what something rated 3500 is doing.

 
AI Leela Chess Zero ID 395 has found a new move in the Nimzo-Indian Defense. Leela, playing as white against chess engine Gull 3, opted for the Normal variation of this opening. What occurred on move 7 however was anything but normal. It was a move typically made only by a beginner chess player. With this move 7 novelty white concedes some material, but its deeper point severely compromises black's kingside so much so that the game is over by move 20. PGN: 1. d4 Nf6 2. c4 e6 3. Nc3 Bb4 4. e3 O-O 5. Bd3 { E47 Nimzo-Indian Defense: Normal Variation, Bishop Attack } b6 6. e4 Bb7 7. e5 Bxg2 8. exf6 g6 9. Bg5 Bxh1 10. Qg4 d5 11. Qh4 Nd7 12. f3 Rb8 13. Kf2 dxc4 14. Be4 c5 15. Bxg6 hxg6 16. Ne4 Bd2 17. Bxd2 cxd4 18. Bg5 Qc7 19. Qh6 Qxh2+ 20. Qxh2 Ne5
 
[edited to remove HTML elements breaking the forum flow -- VP]
 
Marie-AnneLiz
IMBacon a écrit :
barretoff wrote:
IMBacon escreveu:

No idea and don't care.  Engine chess bores me to tears.

I don't think like you do. The neural networks are very interesting, the movements they make during games are often surprising, I think it's really cool.

I think the technology is awesome.  And the fact that someone got a learning machine to go from beginner to beating the best chess engines in 4 hours is impressive to say the least. 

But the reason why watching them play doesn't interest me?  I dont understand the moves.  I have no idea why 2700-2800 GM's make the moves they do.  So I am definitely wont understand what something rated 3500 is doing.

https://www.youtube.com/watch?v=0Hhhpu2_Ugk

DiogenesDue

Unless there is an upside for Google's publicity, there won't be a match.  At best, you'll get another "behind closed doors" match, where they will claim victory for A0 in a match nobody outside the DeepMind team gets to see wink.png.  A0 can hardly be called anything until it plays in a real match overseen by a neutral 3rd party.  Right now it's an unproven curiosity that plays chess under mysterious conditions, like The Turk.

IMKeto
Marie-AnneLiz wrote:
IMBacon a écrit :
barretoff wrote:
IMBacon escreveu:

No idea and don't care.  Engine chess bores me to tears.

I don't think like you do. The neural networks are very interesting, the movements they make during games are often surprising, I think it's really cool.

I think the technology is awesome.  And the fact that someone got a learning machine to go from beginner to beating the best chess engines in 4 hours is impressive to say the least. 

But the reason why watching them play doesn't interest me?  I dont understand the moves.  I have no idea why 2700-2800 GM's make the moves they do.  So I am definitely wont understand what something rated 3500 is doing.

https://www.youtube.com/watch?v=0Hhhpu2_Ugk

Since i dont play anyone rated 3500, this doesnt help me.

Marie-AnneLiz
IMBacon a écrit :
Marie-AnneLiz wrote:
IMBacon a écrit :
barretoff wrote:
IMBacon escreveu:

No idea and don't care.  Engine chess bores me to tears.

I don't think like you do. The neural networks are very interesting, the movements they make during games are often surprising, I think it's really cool.

I think the technology is awesome.  And the fact that someone got a learning machine to go from beginner to beating the best chess engines in 4 hours is impressive to say the least. 

But the reason why watching them play doesn't interest me?  I dont understand the moves.  I have no idea why 2700-2800 GM's make the moves they do.  So I am definitely wont understand what something rated 3500 is doing.

https://www.youtube.com/watch?v=0Hhhpu2_Ugk

Since i dont play anyone rated 3500, this doesnt help me.

 

Do you really need to be or to talk to a PhD in physics to want to know some basic understanding in quantum physics.

What is quantum physics? Put simply, it's the physics that explains how everything works: the best description we have of the nature of the particles that make up matter and the forces with which they interact. Quantum physics underlies how atoms work, and so why chemistry and biology work as they do.

Caesar49bc

I thought Google dismantled Alpha Chess Zero.

IBM famously dismantled Deep Blue after winning against Kasparov. -Although if you look at the hardware back then, I think a top of the line cell phone from 2020 would be superior from a pure hardware point of view.

IMKeto
Marie-AnneLiz wrote:
IMBacon a écrit :
Marie-AnneLiz wrote:
IMBacon a écrit :
barretoff wrote:
IMBacon escreveu:

No idea and don't care.  Engine chess bores me to tears.

I don't think like you do. The neural networks are very interesting, the movements they make during games are often surprising, I think it's really cool.

I think the technology is awesome.  And the fact that someone got a learning machine to go from beginner to beating the best chess engines in 4 hours is impressive to say the least. 

But the reason why watching them play doesn't interest me?  I dont understand the moves.  I have no idea why 2700-2800 GM's make the moves they do.  So I am definitely wont understand what something rated 3500 is doing.

https://www.youtube.com/watch?v=0Hhhpu2_Ugk

Since i dont play anyone rated 3500, this doesnt help me.

 

Do you really need to be or to talk to a PhD in physics to want to know some basic understanding in quantum physics.

What is quantum physics? Put simply, it's the physics that explains how everything works: the best description we have of the nature of the particles that make up matter and the forces with which they interact. Quantum physics underlies how atoms work, and so why chemistry and biology work as they do.

"Do you really need to be or to talk to a PhD in physics to want to know some basic understanding in quantum physics."

Their is a huge difference between some basic understanding of something, and understanding what something rated 800+ points higher than the best humans play.

Besides...all i am saying is that engine matches do not interest me.  They do not interest me because i do not understand the moves.  Thats all.

Marie-AnneLiz
Caesar49bc a écrit :

I thought Google dismantled Alpha Chess Zero.

IBM famously dismantled Deep Blue after winning against Kasparov. -Although if you look at the hardware back then, I think a top of the line cell phone from 2020 would be superior from a pure hardware point of view.

Deep Blue in 1997 was capable of evaluating 200 million positions per second, twice as fast as the 1996 version. In June 1997, Deep Blue was the 259th most powerful supercomputer according to the TOP500 list, achieving 11.38 GFLOPS on the High-Performance LINPACK benchmark.

The Deep Blue chess computer that defeated Kasparov in 1997 would typically search to a depth of between six and eight moves to a maximum of twenty or even more moves in some situations.

congrandolor
Marie-AnneLiz wrote:
IMBacon a écrit :
barretoff wrote:
IMBacon escreveu:

No idea and don't care.  Engine chess bores me to tears.

I don't think like you do. The neural networks are very interesting, the movements they make during games are often surprising, I think it's really cool.

I think the technology is awesome.  And the fact that someone got a learning machine to go from beginner to beating the best chess engines in 4 hours is impressive to say the least. 

But the reason why watching them play doesn't interest me?  I dont understand the moves.  I have no idea why 2700-2800 GM's make the moves they do.  So I am definitely wont understand what something rated 3500 is doing.

https://www.youtube.com/watch?v=0Hhhpu2_Ugk

awesome game! thanks for sharing it

Marie-AnneLiz
IMBacon a écrit :
Marie-AnneLiz wrote:
IMBacon a écrit :
Marie-AnneLiz wrote:
IMBacon a écrit :
barretoff wrote:
IMBacon escreveu:

No idea and don't care.  Engine chess bores me to tears.

I don't think like you do. The neural networks are very interesting, the movements they make during games are often surprising, I think it's really cool.

I think the technology is awesome.  And the fact that someone got a learning machine to go from beginner to beating the best chess engines in 4 hours is impressive to say the least. 

But the reason why watching them play doesn't interest me?  I dont understand the moves.  I have no idea why 2700-2800 GM's make the moves they do.  So I am definitely wont understand what something rated 3500 is doing.

https://www.youtube.com/watch?v=0Hhhpu2_Ugk

Since i dont play anyone rated 3500, this doesnt help me.

 

Do you really need to be or to talk to a PhD in physics to want to know some basic understanding in quantum physics.

What is quantum physics? Put simply, it's the physics that explains how everything works: the best description we have of the nature of the particles that make up matter and the forces with which they interact. Quantum physics underlies how atoms work, and so why chemistry and biology work as they do.

"Do you really need to be or to talk to a PhD in physics to want to know some basic understanding in quantum physics."

Their is a huge difference between some basic understanding of something, and understanding what something rated 800+ points higher than the best humans play.

Besides...all i am saying is that engine matches do not interest me.  They do not interest me because i do not understand the moves.  Thats all.

I meant some Basic understanding of a FEW the new moves by A0 or Leela...

Like this one:

https://www.youtube.com/watch?v=0Hhhpu2_Ugk

I agree with you mostly of course!

Marie-AnneLiz
Caesar49bc a écrit :

I thought Google dismantled Alpha Chess Zero.

IBM famously dismantled Deep Blue after winning against Kasparov. -Although if you look at the hardware back then, I think a top of the line cell phone from 2020 would be superior from a pure hardware point of view.

Deep Blue in 1997 In June 1997, Deep Blue was the 259th most powerful supercomputer according to the TOP500 list, achieving 11.38 GFLOPS on the High-Performance LINPACK benchmark.

Apple 11 in 2018 = 154.9 GFLOPS

Gigaflops is a unit of measurement used to measure the performance of a computer's floating point unit, commonly referred to as the FPU. One gigaflops is one billion (1,000,000,000) FLOPS, or floating point operations, per second.

JamesAgadir

Never. 

ponz111

Alphazero.  but L:eela very good.