End of Era is there: SF is finally beaten!

[CMCanavessi]

Leela about to go +3, SF fans already using excuses like "hardware disadvantage" or "no book"... gimme a break.

[Alexander Schmidt]

Leela about to go +3, SF fans already using excuses like "hardware disadvantage" or "no book"... gimme a break.

The same guys who made fun about the argument, Lc0 would be better without book?

[chrisw]

After game 10, Stockfish was two games ahead. Which obviously means the era of Stockfish continues and LZ is hopelessly weak.
After game 50 something, LZ is two games ahead, so it's the end of the Stockfish era. Bla bla.

With a non-draw-result to games played ratio of about 1 in 5 or whatever it is, the chances of two equal programs each being two games up at some point up in a long game sequence is high enough. Laskos could probably do the maths. It's a great statistical fallacy to cherry pick the point in the sequence you want to stop counting, and declare a result.

[M ANSARI]

There is no doubt that we are in a new Era ... you have to be living under a rock if you don't see that. Lc0 basically outplays SF almost every game. Sure it doesn't win many games it has a big advantage in and will lose to one move tactical blunders because it still has some tactical weaknesses, but in general it just plays much stronger chess than any AB engine. As for people who mention the hardware disparity ... are you kidding me ???? SF is running on a 128 core machine!!! Lc0 is running on a 2080Ti card and another 2080 card. These are GPU's that are designed for gaming but happen to also have some hardware that can do NN. That is like having a CPU run video graphics ... it will work but is not really ideal. I expect the actual hardware that Lc0 uses to become 10x or even 100x or maybe even 1000 or 10,000x more powerful in the very near future as all the big players get onboard. Already Intel plans to release its first NN chip in 2 years and Apple, Microsoft, Google and basically everyone else is getting on the AI band wagon. This is the beginning of a new Era and it really has very little to do with chess or chess engines.

By the way, everyone assumes that SF or other top AB engines are just going to sit idly by as NN engines take over. There is nothing that prevents SF from using NN to enhance its AB engine. It is very obvious (for now at least) that Lc0 suffers in tactics and that SF or other strong AB engines play tactical positions much better. I still don't understand why Lc0 and A0 suffer from this and can basically be blind to one move blunders, but if it is something inherent with how NN works, then Lc0 will also benefit from going hybrid and using some sort of AB sanity check.

[Alexander Schmidt]

After game 10, Stockfish was two games ahead. Which obviously means the era of Stockfish continues and LZ is hopelessly weak.
After game 50 something, LZ is two games ahead, so it's the end of the Stockfish era. Bla bla.

With a non-draw-result to games played ratio of about 1 in 5 or whatever it is, the chances of two equal programs each being two games up at some point up in a long game sequence is high enough. Laskos could probably do the maths. It's a great statistical fallacy to cherry pick the point in the sequence you want to stop counting, and declare a result.

It's not about this one match, and who is a few ELO stronger. It's about new techniques which will give computerchess a huge boost. Soon there will be attempts to use the knowledge of Leela with the experience of 50 years computerchess programming. We are for sure at the beginning (!) of a new Era. I guess at the end chess will be kind of solved and we will have no idea why.

BTW: In chess Leela is not very successful compared to what it is doing in Go. A few years ago all said, a computer will never be able to play well the game of go. Now it is dominating the humans, even moving instantly without search only the best human players can beat it.

It's kind of funny how a few people hate Leela for what it is doing. All will benefit from it.

The only thing we should pay attention about is how reinforced learning will change our life. Soon computers will make decisions which we don't understand. I really hope it will not lead in a world where robots with AI will decide to shoot a human because the AI sees a 51% probability that the human is dangerous.

[George Tsavdaris]

There is no doubt that we are in a new Era ... you have to be living under a rock if you don't see that. Lc0 basically outplays SF almost every game. Sure it doesn't win many games it has a big advantage in and will lose to one move tactical blunders because it still has some tactical weaknesses, but in general it just plays much stronger chess than any AB engine. As for people who mention the hardware disparity ... are you kidding me ???? SF is running on a 128 core machine!!! Lc0 is running on a 2080Ti card and another 2080 card.

Stockfish is running on 43 cores (2 x Intel Xeon E5 2699 v4 @ 2.8 GHz Motherboard: Supermicro X10DRL-i RAM: 64 GB DDR4 ECC SSD: Crucial CT250M500 240 GB) but still your point stands.

[Graham Banks]

How can one know whether the engines are playing on equal terms?
How can you compare CPU and GPU and state you've found a balance?

[M ANSARI]

I think the hardware is irrelevant as on many winning ideas even the 170+ thread SF was totally unaware of any danger until it was too late. The result of this TCEC is really not important as it is not testing what engine is strongest, but rather it is an engine match created to make it as entertaining as possible. Looking at the result of the games, you can very quickly see that the book opening can easily dictate which side would win. The idea was to throw each engine in a position that it is not comfortable with and see if it can survive, or in the case of the attacking engine if it can find a winning plan. This should not be confused with trying to find the strongest engine, although of course since games are reverse colors, it does give a very good indication of an engine strength over another.

I think GPU's being used for chess will be only a temporary thing as for chess a much better solution would be to use cards that are specifically designed to do NN and nothing to do with graphics. This will probably become very common very soon and some kind of hardware equation needs to be agreed upon if AB vs NN engine matches will take place. Most likely SF and Komodo and Houdini will become hybrid engines that use both NN and AB and thus will make full use of the CPU as well as the GPU.

[corres]

How can one know whether the engines are playing on equal terms?
How can you compare CPU and GPU and state you've found a balance?

Nohow.
There are only subjective methods to make comparison.
Moreover AB engines are made basically for playing middle games and endgames.
An AB engines will be banished from a competition if it has a built-in opening book.
An NN type engines has definitively a built-in opening + middle game + endgame book.
AB engines have their own benefit and their own drawback and NN type engines have these too.
It is obvious the development of hardware and software gives more opportunity for enhancing
the chess power of the NN type engines than for an AB type engines.
But the problem of chess can not be solved by an NN type engine either.

[Graham Banks]

The result of this TCEC is really not important as it is not testing what engine is strongest, but rather it is an engine match created to make it as entertaining as possible.

Makes sense, and it sure is close and entertaining.
However, many will claim otherwise.