Next step for SF NNUE: Chessbase engine to buy?

[Jouni]

There has been one so far. Houdini 6 = Stockfish 8.5 .

[smatovic]

There has been one so far. Houdini 6 = Stockfish 8.5 .

Hehe, I bet on "Fat Fish"

--
Srdja

[Ovyron]

There has been one so far. Houdini 6 = Stockfish 8.5 .

Hehe, I bet on "Fat Fish"

--
Srdja

No, the point isn't the Fat part, but the Fritz part.

So my bet goes for Freak Fritz! The strongest Fritz to date, and the most commercially successful open source engine!

[Thomas Lagershausen]

Come on.

Business people making business stuff.

[Jouni]

I think SF + NNUE means the end of commercial engines. End of an era.

[Raphexon]

I think SF + NNUE means the end of commercial engines. End of an era.

Mark Lefler has reached out to NNUE dev...

If he and the Komodo team find a promising nnue architecture and quickly start training they may be able to get close to SF again, despite inferior search.
The current top nets are produced by Sergio who doesn't have that much hardware, so if they strike gold they can produce nets of superior quality than what SF is currently using.

It's just an "if" though

[lucasart]

I think SF + NNUE means the end of commercial engines. End of an era.

The end was 16 years ago already, when Fruit started challenging the Shredder supremacy. It's been an uphill battle for commercial chess engines ever since.

The model of distributed developpement is unbeatable. SF has access to all the brain power and computing power that is interested in it. And the stronger it gets, the more it accretes brain power and computing power. It's basically a black hole. Commercial engine developpers are hopelessly trying to exceed the speed of light to escape the SF black hole. Impossible: SF event horizon is behind already.

Developping a chess engine on one's own is still fun (for some). But doing so competitively is a hopeless pursuit.

[syzygy]

I think SF + NNUE means the end of commercial engines. End of an era.

Mark Lefler has reached out to NNUE dev...

If he and the Komodo team find a promising nnue architecture and quickly start training they may be able to get close to SF again, despite inferior search.
The current top nets are produced by Sergio who doesn't have that much hardware, so if they strike gold they can produce nets of superior quality than what SF is currently using.

It's just an "if" though

Are the NNUE developers willing to license the code? (Is it fully theirs to begin with? Probably it is, but I have no idea.)

[syzygy]

Developping a chess engine on one's own is still fun (for some). But doing so competitively is a hopeless pursuit.

Well, if a commercial developer can manage to keep a new idea like NNUE for themselves, there are still possibilities.

But I agree that the chances that the next new idea will be developed on top of Stockfish and not be kept secret are much greater than the chances that it will come from one of the few small groups of commercial developers.