Leela with massive hardware will be participating in ICGA tournment

Sort:
SmyslovFan

I don't think that will ever happen, @btickler. A0 has accomplished its goal in terms of chess accomplishment. It's already moved on. Only nerdy chess players are interested in seeing whether it's actually as good as it claims to be. 

drmrboss

Here is the link for live-games.

 null

http://view.livechesscloud.com/0939e4c8-636e-41c0-98ac-c87bd0cc65e6

http://view.livechesscloud.com/3d47fbee-2b8b-4154-b004-42d084b68982

 

AnhVanT

Will Leela be able to perfectly play like any GM in the past by learning thousands of their game? 

congrandolor
SmyslovFan wrote:

I don't think that will ever happen, @btickler. A0 has accomplished its goal in terms of chess accomplishment. It's already moved on. Only nerdy chess players are interested in seeing whether it's actually as good as it claims to be. 

Bill Gates is a nerdy chess fan, maybe sb could persuade him to afford an A0 vs SF rematch

drmrboss
AnhVanT wrote:

Will Leela be able to perfectly play like any GM in the past by learning thousands of their game? 

There is no perfect play in chess, except table-bases.

Leela on a decent GPU hardware is in top 10 chess engine level. (way beyond human GM level)

Leela knocked out top 10 rated Ethereal.

null

drmrboss
btickler wrote:

Hopefully, A0 will play

 

a real match someday.

null

This is theoretical estimate of A0 rating vs stockfish on different hardwares. In fact stockfish on 386 thread could be much stronger than A0.

drmrboss

The first game of Leela (white)vs Komodo (black), Leela lost on time on TB draw endgame. 

 

Preggo_Basashi
drmrboss wrote:

The first game of Leela (white)vs Komodo (black), Leela lost on time on TB draw endgame. 

 

 

Did it have access to the TB? I guess it ran out of time before it realized?

superchessmachine

I do not know Preggo. Computers are very fast. It could have found out in a millisecond if it tried I think

superchessmachine

Analyzing that game with chess.com stockfish maximum analysis level took 40 minutes.

DiogenesDue
SmyslovFan wrote:

I don't think that will ever happen, @btickler. A0 has accomplished its goal in terms of chess accomplishment. It's already moved on. Only nerdy chess players are interested in seeing whether it's actually as good as it claims to be. 

That's because only "nerdy" chess players and some "nerdy" programmers are aware of the fraud that was perpetrated...

One could just as easily say that nobody is interested in Einstein, because they can only understand Newton's theories without taking a class on relativity.  Einstein is still relevant, and the majority of humanity's fostering of their own ignorance is not, as per usual.  News outlets will copy and spread anything that sounds remotely interesting without any scrutiny.  The Google press releases were designed to take advantage of that, and that is what they succeeded in.  Any truly real "chess accomplishments" achieved were kept in that back room...along with lots of other things, like how many times the DeepMind team could not get the crushing results they wanted until they finally figured out the right combination of hamstringing settings that would cripple Stockfish just enough...

drmrboss
btickler wrote:
SmyslovFan wrote:

I don't think that will ever happen, @btickler. A0 has accomplished its goal in terms of chess accomplishment. It's already moved on. Only nerdy chess players are interested in seeing whether it's actually as good as it claims to be. 

That's because only "nerdy" chess players and some "nerdy" programmers are aware of the fraud that was perpetrated...

One could just as easily say that nobody is interested in Einstein, because they can only understand Newton's theories without taking a class on relativity.  Einstein is still relevant, and the majority of humanity's fostering of their own ignorance is not, as per usual.  News outlets will copy and spread anything that sounds remotely interesting without any scrutiny.  The Google press releases were designed to take advantage of that, and that is what they succeeded in.  Any truly real "chess accomplishments" achieved were kept in that back room...along with lots of other things, like how many times the DeepMind team could not get the crushing results they wanted until they finally figured out the right combination of hamstringing settings that would cripple Stockfish just enough...

Nothing nerdy! No need to know technical details.

 

For example a power supply of "5W" is crippling to a computer performance whereas this number become too different from common use. (Just for general awareness such as a cell phone uses "5W and a computer use "500 W")

And also this is similar to "1G" memory to current computers. 

4 cores use 8 or 16 G

So, 32 cores would need, 64 G or 128 G.

drmrboss

Second rould of leela, draw vs boot!!

 

drmrboss

Third round of leela, lose to Gingko

 

drmrboss
Preggo_Basashi wrote:
drmrboss wrote:

The first game of Leela (white)vs Komodo (black), Leela lost on time on TB draw endgame. 

 

 

Did it have access to the TB? I guess it ran out of time before it realized?

The operators manually enter moves into the chessboard!!

Leela operator could not manually enter the moves into chessboard on time, so Leela lose. (non sense outdated rules since 1970, they never update the rules for 50 years) 

A lot of people blame on those tournment rules! 

IpswichMatt
btickler wrote:

Any truly real "chess accomplishments" achieved were kept in that back room...along with lots of other things, like how many times the DeepMind team could not get the crushing results they wanted until they finally figured out the right combination of hamstringing settings that would cripple Stockfish just enough...

Gosh, do you really think they would do such a thing?

I'm shocked to my core

SmyslovFan

One of the points behind Leela and AlphaZero is that they have no access to table bases or opening books. They create everything for themselves. So no, they don't have a tablebase unless they create it themselves in the run-up to the tournament.