Id 234 is out , nice rating jump compared to previous ids. But i won on time, she was slow. I see different openings and improved lines.
Leela Zero( A Neural Network engine similar to Alpha Zero)
Sort:
Oh, an oldie, most popular chess engine the King aka Chessmaster. is down.
Recent achievements of Leela.
1. Beat Stockfish 1.3
2. Beat Komodo 1.00
3. Beat the King 3.50
4. Beat Rybka 2.3
Currently testing but still losing vs
1. Houdini 1.5
2. Stockfish 3.0
3. Rybka 4.1
Enthusiasts, myself as well, expect that Leela can beat latest Stockfish in decent GPU like 1060GTX within 6 months.
[COMMENT DELETED]
https://code.facebook.com/applied-machine-learning/
https://news.slashdot.org/story/18/05/02/2224251/facebooks-open-source-go-bot-can-now-beat-professional-players
Facebook is also doing machine learning and they claimed that their AI beat 3rd best GO AI, "Leela Zero", which was controversial though.
Anyway, there could be Artificial Intelligence wars/competition in gaming industry and other Scientific fields. 
?Best Go AI
1. Alpha Go Zero
2. ?Elf
3. ? Leela Zero
? Best chess AI
1. Alpha Zero
2. ? Leela Chess Zero
3. ? ? no info about facebook AI on chess
Here is cuDNN version of Leela ,which is about 4 times faster, Time to to test vs stockfish.
As I am just amateur with very limited level of C++, it was very difficult for me to understand how things work. 
Steps 1.
1. Upgrade latest Nvidia driver
2. Download and install visual studio 2015 (not sure this is a requitement for all machines), visual studio 2017 did not work for me
3. Download CUDA 9.0 (9.1 did not work for me),
Install CUDA in optional mode and uncheck visual studio integration on installation
3. Download cudNN 9.0
4. Unzip and paste the necessary files exactly according to nvidia instruction.
https://ibb.co/eGFgon
5. copy binaries from this link https://crem.xyz/lc0/ and unzip into ur installed directory. eg. C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v9.0\bin
6. Download and install the latest version of leela into the same directory as in step 5. (they two must be together in the same directory as cuDNN version of leela needs some dll files , I think)
7. Download weights and install in Arena.
https://ibb.co/myORon
The speed increased about 4 times, Wow, worth doing. 
drmrboss wrote:
I had never heard of ELF before, but it sounds like it can be used for games other than Go. But that would have been more interesting a few months ago before Leela was forked. Nevertheless, it is nice to see such open source projects everywhere. Can be only good.
A very good tactical game from stockfish in TCEC season 7, stockfish sacrificed a rook for two pawns. I am not sure Leela can stand in such tactical sharp lines (note. stockfish analysed another sf games as multiple inaccurancies).
https://lichess.org/394ubpBA#1
Leela is using Monte Carlo tree search where she plays some random games from the current position. Based on the result of these games, it chooses the move. This is my understanding so far.
But how the training games it plays helps in this process. How she uses the training game during search.
Does she use coefficient while selecting random moves from a position to play over the games in Monte Carlo search? And the coefficients is like strategy in chess. .Right?
Credit, Ali.
importantheart wrote:
Who's Ali?
A TCEC fan. You can see his Artworks,
https://plus.google.com/+AliTCECS11
Credit Ali!
Comparing the speed between openCL version of Leela (1knps) vs cuDNN version (4knps) ,
Although 4x speed faster, testers say cuDNN leela is only about 90 elo stronger. ( expected increase in elo =200 elo)
Xd
A position that embarrass stockfish.
Leela see the solution immediately and some people say stockfish take up to 30 minutes
Analysis at 3 seconds and 1 min by both engines.(my brainfish dont see after 1 min) 
Solution. Rd1-d8
Nice to see. I thought 230 would be really strong, but managed to expliot a tactical blunder in my opening and then checkmated her. Time control 0 plus 1 (3sec plus 1 for whole game)
https://lichess.org/2nBXciROdvCc