The future of computer chess

[Christopher Martin]

Friends of Talkchess, ladies and gentlemen, I am writing to you to share an unusual calculation. I follow very closely the inquiries of genius.

ChatGPT, twinned with multimodal agents, propositional logic and conceptual anchoring, is capable of surfing waves as high as 4500 elo. Who would have thought it. The heritage of light and expertise does not live in the fish riders. Scientific advancement works its way through the narrow minds of obsolete schoolchildren. I can only laugh, yes, I laugh. Ha, ha, ha, ha, ha! exclaim my judicious and cracked lips. Ho, ho, ho, ho, ho! mumble my affiliated teeth. The future is today, no matter who it may concern.

[towforce]

Clickbait. LLMs are extremely good at what they do in comparison to anything publicly available before November 2022 - but what they do is not top level chess.

[towforce]

Clickbait. LLMs are extremely good at what they do in comparison to anything publicly available before November 2022 - but what they do is not top level chess.

I think what you should have said is that advancement in knowledge technologies will make the Stockfish project obsolete in a few years. That I could agree with. LLMs won't do that, though.

[Dann Corbit]

AI engines are being used to improve algorithms successfully now.
I guess it is a matter of time before all programming is done by AI agents.

[towforce]

AI engines are being used to improve algorithms successfully now.
I guess it is a matter of time before all programming is done by AI agents.

That's not wrong - but I'll offer "a different perspective on the same thing".

It's part of a long-term trend that's been going on ever since the first high level programming language (Fortran) was created in 1954: the move from "imperative" coding (telling the computer exactly what to do step by step) to "declarative programming" (link) in which you tell the computer what you want.

It's conceivable to me that soon we'll need a new paradigm for "the computer works out what you want and gives it to you without you having to ask".

[mclane]

?
I think you are a little to strict. Fortran was the mistake? The mistake was to rely on hardware instead of software.

The better the machines the less progress.

With 6502 5mhz kittinger, schroeder made 1900-2000 elo.

The university guys always had massive hardware but minor software.
In 1986 Ed schroeder nearly won the open computerchess championship against the mainframes.


I know stockfish has much more elo. But stockfish has not only more then 5 mhz, it has also more then 32 kb size of code and more then 8 kb RAM. Also more then one cpu and also NN support.

This is almost all hardware shit.

Software beats hardware.
And if you rely on hardware… you lose IMO.

Frans morschs engines can run on very low resources.
Mark taylor wrote an engine on a 4Bit cpu !
Thats the hardware we all had many years as Calculators at school.



So the true pioneers were Kittinger, schroeder, rathsmann, nitsche+henne and broughton etc etc

They made so much elo out of almost nothing.


They will be seen in 200 or more years like galilei or newton or descartes or einstein.

My heroes are those pioneers of computerchess. Richard lang. Martin bryant. Marty Hirsch and don dailey.
I visit the graves of bobby fischer, david broughton and don dailey, wolfgang delmare.

[towforce]

With 6502 5mhz kittinger, schroeder made 1900-2000 elo.

I know stockfish has much more elo. But stockfish has not only more then 5 mhz, it has also more then 32 kb size of code and more then 8 kb RAM. Also more then one cpu and also NN support.

Yes - it would be difficult to to overstate how bad the 6502 and the other great 8-bit CPUs like Z80 and 8088 were compared to today's CPUs (or more properly SOCs): the following might happen in an instruction cycle:

* execute an instruction

* write some data to memory (which, apart from a tiny number of registers, was all on separate chips!)

* fetch the next instruction (from memory on another chip)

All the coding had to be done in assembly language, and the code also had to monitor the buttons on the dedicated chess computer for user interaction.

There were a lot of ways time had to be used doing things other than processing chess code with an 8-bit CPU with no memory apart from a few registers.

Roughly speaking, a linear increase in elo has required an exponential increase in hardware and software resources.

[smatovic]

[...]
Roughly speaking, a linear increase in elo has required an exponential increase in hardware and software resources.

Comparing Chess Engines over History or Architectures - Elo / (Transistorcount*Frequency)
viewtopic.php?p=938469

--
Srdja

[towforce]

[...]
Roughly speaking, a linear increase in elo has required an exponential increase in hardware and software resources.

Comparing Chess Engines over History or Architectures - Elo / (Transistorcount*Frequency)
viewtopic.php?p=938469

--
Srdja

Good find! Thank you for archiving and retrieving that.

[mclane]

It is always said that there was programming progress. But this is not really true IMO. Mainly the hardware and the resources made it possible to use more size, more ram and more cpu power and THAT made the progress in programming possible.

Stockfish would not work on 6502 because the engine is too big and needs ram and etc.

Todays chess programmers are invited to make stronger engines on those old platforms to prove me wrong.

If someone beats morsch, Schroeder or kittinger on z80 or 6502 because „the software made progress“ please demonstrate it with your engine on z80 or 6502!!

Isnt this a good scientific contest ? Beat the Ti99/4a chess engine.
Beat the intellivision chess with your own engine on that platform.


Show us that there really was progress made in software in those 40 years.

Beat cyrus II or Sargon or colossus chess4 on the C64 or the on the acorn bbc B micro computer . The zx spectrum or the oric atmos, the Thomson To8d or the amstrad cpc464.

Show us the superiority of the software progress.