Thoughts about a new "Man vs Machine" type of games

It's well known that the top engines have become untouchable by any human player, so the interest in human-computer matches has mostly waned. There have been some efforts to level the playing field but I don't agree with them.

Giving extra moves or giving up pieces devalues the match IMO. It isn't chess if one side doesn't have all the pieces heh... also using an engine for analysis also defeats the purpose. I mean these may be interesting but they are not in the spirit of "Man vs Machine"

I propose the creation of "Reference" hardware, a computer with a specific CPU and memory, for example a dualcore ARM processor and 256mb RAM, on which all chess engines will run now and in the future. The computing power should be such that in tournament time controls and with engine pondering turned on, today's top engines (Stockfish and Komodo) would be at around 2800 elo strength. So we can see matches of GMs against "Reference-Houdini 4", "Reference-Critter"  etc.

Also, extra points if the computer is inside a robot which will move its own pieces by itself and have voice playback for calling checks and mates

Or maybe the board can be magnetic and the pieces will just slide in place by themselves.

Makes no sense.  By pure calculation computers will always be unbeatable by the strongest chess champion so what's the point?  The only match I would be interested in would be a world class Correspondence Grandmaster with several days to make a move against a world class program.  For some reason that has never happened.

High time the Mechanical Turk made a comeback armed with Komodo rather than a tiny grandmaster.

and trash talking...Lots of trash talking...It would be a smash hit...

Actually, I can't abide by trash talk, but that seems to be what the public wants these days.

Well, that's why this will have a very modest CPU, so that its computational power is limited, so the chess engine will have to manage its time slices carefully.

And why are you proposing centaur chess basically? That's way more boring than what I had in mind, heh.

I have no interest in centaur chess.  I would like to see a Correspondence Grandmaster with no computer help go against an engine with no time controls.  I remember Hans Berliner who is a Correspondence Grandmaster stating one time that Over the Board Grandmasters were several hundred points weaker in play than Correspondence players who had days to make a single move. 

I can't find it now but I read about a female correspondence player in the 19th century who saw mate in something like 90 moves or something.  I don't remember exactly how many moves it was but it was unbelievable.

In short I would like to see if the strongest engines could outplay a human in positional play where the human had all the time he wanted to calculate a move against the computer. 

And we will have to take the word of the player that in those days he did not use a chess engine. Because chess players are honest

I see where you're coming from but I want to see something way more ... marketable. To get some hype going! And to see some deep games live.

Until the late 90's no computer stood a chance against the top chess grandmasters so they couldn't cheat.  If what Hans Berliner said was true (and it sounds reasonable) then a Correspondence Grandmaster with several days to make a single move would be playing at Stockfish level. 

I have to agree with what someone said about the idea making no sense. To standardize a computer hardware to match humans there has to exist a measurement and definition of human thought process plus there will be very strong individuals who may distort all the results. 

In my opinion there are only two options. One is not worrying about how strong computers are and keep competing against human players and using computers for analysis only. 

The other option is getting rid of chess as a skill medium and as the new AI and hardware get developed, use their decision making process and knowledge about their hardware limitations to create a new game that is specifically designed to challenge computers and which is accessible to humans. 

I just want to give an example of a game that computer AI doesn't grasp well and which, when human players play vs AI the AI is always set to be able to cheat. I see that all the time in turn based 4x Games. Most of you must know what those games are but it refers to games like Civilization and many of its clones and imitator. Civ itself isn't even the best example of a strong and well balanced 4x game, but it is best known. There are many games like that including other Sid Meyer games, Heroes of Might and Maguc series, warlock games, and many more. I've only played Civ games, Fallen Enchantress, Warlock, and some less known ones. 

In all of these games AI sucks and cheats. It is a widely known fact. People want to play vs AI and are often upset that the AI is only "harder" because it is secretly given knowledge of your positions and decisions as well as a resource handicap in favor of AI. 

These 4X games can be complex proportionate to the map size and the map sizes of these games can be absolutely huge and I am sure complexity surpasses that of Go. 

These se games always change and get updates, however, so they are far from being standardized. But I think training AI to be very good at a good 4X game will ever more difficult than training the AI to be good at Go. There are just many variables that are dependent on chance and invisible dive rolls as well as random events. 

This hardware standardization is just to keep the engines' strength at around a human GM's strength so we can have some nice showcase matches, nothing more. Think of it as a way to keep it competitive without resorting to weird "odds" being given.

Also, it doesn't have to be the top engines being played, there are other engines that can play interesting and aggressive chess, too.

I would like to see a tournament with 8 GMs and 8 engines taking part