Proposal: This forum should actively support LCZero

[JohnS]

LCZero is a promising new paradigm for developing a super strong open-source engine. The engine is weak at the moment but is getting stronger all the time.

What the project needs is computing power to generate training games. If most members of the forum (around 1600) contributed a bit of their computing power, progress would speed up considerably.

Carlos Canavessi has provided a ready to run system (http://www.talkchess.com/forum/viewtopi ... =&start=30) which makes it very easy to help (thanks Carlos). I was able to install the files and start contributing in five minutes. No skill is required!

So let's all pitch in and speed up the development.

Finally, if you don't like the proposal please just ignore the request. I don't want this thread to be highjacked by people who don't like LCZero.

[Ozymandias]

Finally, if you don't like the proposal please just ignore the request. I don't want this thread to be highjacked by people who don't like LCZero.

Good luck with that. And good luck with LCZero too, although I remain skeptical.

[Jhoravi]

Carlos Canavessi has provided a ready to run system (http://www.talkchess.com/forum/viewtopi ... =&start=30) which makes it very easy to help (thanks Carlos). I was able to install the files and start contributing in five minutes. No skill is required!

I tried it and it worked like a charm but IMO the process is still not friendly enough to motivate my chess playing friends. I wish this is some form of an app or webpage with a simple push of a button then I'll convince them again.

[peter]

Hi!

I wish this is some form of an app or webpage with a simple push of a button then I'll convince them again.

As for my pov. it's less a matter of easy access only, rather one of any kind of data- security too.

Opening up my computer with PPP to a network I don't know nothing about seems somewhat problematic to me. Could that be made somewhat transparent maybe?

[Dann Corbit]

It is an interesting experiment.
But you will never get consensus on anything.

[Werewolf]

I think this would be good to support, though I do have security concerns

[Jesse Gersenson]

Hi!

I wish this is some form of an app or webpage with a simple push of a button then I'll convince them again.

As for my pov. it's less a matter of easy access only, rather one of any kind of data- security too.

Opening up my computer with PPP to a network I don't know nothing about seems somewhat problematic to me. Could that be made somewhat transparent maybe?

Downloading files from a very dubious site, mediafire, then running closed source exe's with network access is a terrible idea.

The transparent solution is:
1. provide the source files for the client
2. post data to a webserver

I built a few such prototype systems several years ago with Don Dailey. The client would go to web server and pull down the jobs it should run. Jobs would run and be stored in either a flat file or a sqlite db. When a job finished it would be uploaded to the web server.

Storing the results locally allowed users to run off-line, if they wanted.

Also, the program would check for updates - and current versions of the engines would be downloaded.

[Jesse Gersenson]

It seems the client source is available:
https://groups.google.com/forum/#!topic ... R0MGWQ93UY

[Jesse Gersenson]

https://github.com/glinscott/leela-chess/releases

[CMCanavessi]

Hi!

I wish this is some form of an app or webpage with a simple push of a button then I'll convince them again.

As for my pov. it's less a matter of easy access only, rather one of any kind of data- security too.

Opening up my computer with PPP to a network I don't know nothing about seems somewhat problematic to me. Could that be made somewhat transparent maybe?

Downloading files from a very dubious site, mediafire, then running closed source exe's with network access is a terrible idea.

The transparent solution is:
1. provide the source files for the client
2. post data to a webserver

I built a few such prototype systems several years ago with Don Dailey. The client would go to web server and pull down the jobs it should run. Jobs would run and be stored in either a flat file or a sqlite db. When a job finished it would be uploaded to the web server.

Storing the results locally allowed users to run off-line, if they wanted.

Also, the program would check for updates - and current versions of the engines would be downloaded.

Where did you get the (wrong) idea that its a closed source project? Please dont spread fud like that. You can go to github download sources, compile them yourself and run everything "safe". The package I uploaded to mediafire was just to help users get going in an easier way. It is in no way official or anything like that. Actually its already outdated (it still works ok though) as progress is being made constantly.