Leela (lack of) endgame technique?

[bob]

LC0 (formerly LCZero), a new approach in computerchess to play the best chess by using a "Neural Network" instead of prunning moves away by "alpha-beta-prunning" (as current engines are based on still momentarely as SF/Houdini/Komodo are), is more or less 1 year among us and performs increadable ca 3300 elo @ CCCC1...
can u guys remember how long it took us to make the a-b-prunning approach working that well that the engines were considered GM-strengh ?
it took many....., many years....
now, with this (absolutely) amazing new concept, this "thing" (no influence by human being...it just plays the game vs itself again and again and noone has ever sayed a single thing about chess exept the rules) plays chess already (lets say: it trained chess for around 3-4 months and now is getting restarted and restarted for reasons i dont have a look for) on a level that NO human is competitive with (probably!?...still no competition was brought into life against a GM)...if u like to have a chess-engine from scratch that is superior to any contender and is build within a few months then leave solarsystem, diss the programmers and wait until it falls into your lap while trolling about their inaccuracies... but for real (imo):
in a few years at the latest i think this will be the new monument of chessprogramming (how far away from the top seats is LC0 currently, 200 elo ? lol)
does it lack of knowledge?...yes oc as a-b-engines...will it lack of knowledge in the future?...yes oc as a-b-engines(in my opinion they will stay important for tactical discourses)....will this approach overshadow the former programmers approach of alpha-beta-cutting...yes but thats normal with more efficient ways to walk...will it play ever perfect chess?...yes oc, with 32 men tablebases

I wish Robert Hyatt were still active on Talkchess, I would love to hear his take on LC0.

I'm still around. My "take" is that it looks promising. I did some ANN stuff in the middle 80's, working on a tracking algorithm for the air force. IE a complex radar image, where got two consecutive images and had to connect everything on the first image to its counter-part on the second. Worked OK, but the problem was "speed".

This has reminded me of the early days of chess where suddenly, chess 4.0 came on the scene with a true full-width search. It became viable when CDC started to release their new cyber line of super-computers. It looks like we are at that point with ANN programming thanks to the massively parallel GPU-type advancements. So it looks promising. But it does seem to be missing basic tactical problems that will continue to hurt it until they are (somehow) fixed. Some of this will be resolved as new hardware makes larger and larger nets usable. But even without this hardware, it is certainly impressive...

[Ovyron]

Bob just called lc0, a project that leaves Crafty in the dust and that under certain conditions wipes the floor with Stockfish (so people with the right hardware have lc0 clearly as best choice), "promising"...

[bob]

Bob just called lc0, a project that leaves Crafty in the dust and that under certain conditions wipes the floor with Stockfish (so people with the right hardware have lc0 clearly as best choice), "promising"...

It has a ways to go. Its tactics leave a lot to be desired. But more hardware speed will eat into that problem.

BTW what does Crafty have to do with this? Just a mild insult or something. How's your program doing, again?

[MikeB]

Bob just called lc0, a project that leaves Crafty in the dust and that under certain conditions wipes the floor with Stockfish (so people with the right hardware have lc0 clearly as best choice), "promising"...

not nice ... Bob did say "...it is certainly impressive..."

[Ovyron]

BTW what does Crafty have to do with this?

Is Crafty promising or not? I mean, because of the exponential nature of chess, ALL computer chess software "has a ways to go." Stockfish leaves a lot to be desired when compared to Stockfish 13 from the future...

How's your program doing, again?

Apparently I have nothing because the community has decided that it's not fine to just copy and paste code from others, and it has conflicting licences from the code I've "stolen" from many others, because I found silly to have to type things "my way" when there's out there perfectly written code that works.

But for the record, just some 400 elo behind Crafty.

Leela and Neural Network engines are the most amazing thing that has happened in computer chess, mainly because it's something that works and it's as good as the best without relying on any of the techniques that have been praised and reused in the history of programming (and EVERYONE wanting to code from scratch has to use them. To stay competitive you can't do without Null moves...)

Except Leela, not using anything from chess code from before, the engine just trains itself, nobody even knows what her internal parameters are doing.

I think it's not fine to measure her success by how much she could improve. "Promising" is an adjective that would have been fair a year ago. Today, the promises have been already fulfilled (passing the top A/B engine), the rest is not unlike all the other computer chess programs, because we're nowhere near the ceiling, yet.

[OliverBr]

Bob just called lc0, a project that leaves Crafty in the dust and that under certain conditions wipes the floor with Stockfish (so people with the right hardware have lc0 clearly as best choice), "promising"...

It's not even difficult: Get yourself a decent GPU (nVidia ist better because of CUDA), like the mighty GTX1080Ti and there you go. *

(*PS On MacOSX you cannot use a OS later than High Sierra, because Apple and nVidia went different ways)

[Ovyron]

It's not even difficult

Or it can be. My hardware can't be upgraded beyond 4GB RAM, high depth Leela requires big RAM to function, so basically I need a new computer with more RAM for it

What I'm hoping for is for the CPU version to catch up, or if they make one that can use the instructions of DirectX for Windows 8.1 then Leela would be on its way to catch Stockfish on my hardware (which is just 2600 kn/s Stockfish, an easy prey.)

[Dann Corbit]

It's not even difficult

Or it can be. My hardware can't be upgraded beyond 4GB RAM, high depth Leela requires big RAM to function, so basically I need a new computer with more RAM for it

What I'm hoping for is for the CPU version to catch up, or if they make one that can use the instructions of DirectX for Windows 8.1 then Leela would be on its way to catch Stockfish on my hardware (which is just 2600 kn/s Stockfish, an easy prey.)

The CPU version will never catch up. The GPUs are orders of magnitude more powerful than the CPUs. Running LC0 on the CPU is like taking a great chess player and giving him a pre-frontal lobotomy and then asking him for advice.

[Ovyron]

The CPU version will never catch up.

I didn't mean Leela on CPU catching up to Leela on GPU (which can't happen.) I meant Leela on CPU catching up to Stockfish on CPU (which could be a matter of time.)

[MikeB]

It's not even difficult

Or it can be. My hardware can't be upgraded beyond 4GB RAM, high depth Leela requires big RAM to function, so basically I need a new computer with more RAM for it

What I'm hoping for is for the CPU version to catch up, or if they make one that can use the instructions of DirectX for Windows 8.1 then Leela would be on its way to catch Stockfish on my hardware (which is just 2600 kn/s Stockfish, an easy prey.)

The CPU version will never catch up. The GPUs are orders of magnitude more powerful than the CPUs. Running LC0 on the CPU is like taking a great chess player and giving him a pre-frontal lobotomy and then asking him for advice.

Awe come on it's not that bad.... It's more like taking a World Champion at his prime and then having him play himself after he turns 100.