Perfect game ?

[Matty]

It may not have been a commercial success and I did describe this regarding AI on the blitz forum a year or two ago but my old Android game 'Star Dancer' has a learning AI that I had to lessen its intelligence to make it more fun for new users.

It's a very different technique of AI that is based heavily on the style of interface the game uses to play it:

Game play : The main decision a player makes is a series of sliders that determine how their ships fly.

What the AI does is (and I'm proud of its learning ability)

When a human player plays, and wins the AI takes the settings that player uses, compares the total wins with a league table and then for other players spits out the same or similar (slight random mods sometimes) to use against other players.

Over time the AI became far too hard and I could never beat it...so I had to tweak it so that it introduced a few more random elements that caused it to make more odd choices from time to time that made it a lesser opponent..introduce some flaws to its logic.

[ENAY]

I miss Angry Joe's videos. These days he does so many other crappy filler content like Movie reviews and reaction videos that he makes less review videos.

[Steve Elliott]

AI is dumb as fu*k, and not evolved at all...

Not true, google DeepMind (now bought by google for $620 million).  It learnt chess from scratch in hours then beat the current best chess engine in the world - learnt it from scratch with only the rules of chess!  Beat the best human player in the world at go (a much harder task) and before that learnt to play a ton of Atari 2600 games with the only goal to improve it's score.  And even though the games varied, the system didn't - it just learnt to play them all perfectly, no matter what the game required to get a best score.

The point is it's a general purpose learning system that Demis Hassabis hopes will help in solving man's greatest problems.

https://www.youtube.com/watch?v=rbsqaJwpu6A

[Naughty Alien]

..chess rules are linear rules and each doesn't quantify any different output value as they are all same (trough different, predetermined template, /figures moves/)..its literally pure brute force filing machine, repetitive task, for which they are in essence, designed to do after all, just like any other computer...i see no AI worth mentioning about that at all..pure brute force and linear pattern recognition..nothing really spectacular

[Steve Elliott]

Google wouldn't pay $620 million dollars for a chess program lol.  There is far more to the technology than that.

Go is a far tougher task to conquer (and it has, way ahead of predictions).  If you think you can solve go with brute force you are talking bollocks - especially as the tech learnt without human interaction!  These are the rules, go away and play better than anything on the planet.  It did it that for chess in 4 hours.  The fact the tech has taught the undefeated (over many years) go master new ways to play a game centuries old is a testimony to it's usefulness.  Humans learning from machines.  It's not just aimed at games, you're missing the point.

Pure brute force and linear pattern recognition, so elements of human traits then?  Except the computer does it far more efficiently and could eventually solve the major problems of today as the system develops, because it's not bogged down in 'current (lack of) wisdom'.  We will see, but to dismiss a tech worth $620 million shows some kind of arrogance to dismiss it out-of-hand.

[Naughty Alien]

..620 mil USD is reference point ?? For chess program? Thats  what you considered as relevant ? Google payed 3 billion usd for company famous for its smart thermostat that learns users' energy preferences, and automatically adjust the temperature to save energy..something which does for good 10 years over here where im in plenty of controllers as default..whole thing in parts doesn't cost more than 100 bucks in material if you do it yourself..

So, to throw google pay as a reference of how valid is project is really wrong. Chess programs are in terms of 'learning' skills/patterns, easier one to deal with, and whole game is right up to the alley of binary state machines as computers as we know are. Thats not an AI, its extremely streamlined/dedicated task and doesnt go anywhere beyond that. There is more AI, if you will, in airplane flight computer, than this chess thing.Fact that such chess program can take out human competitor doesnt mean anything in terms of intelligence. Its brute force/sorting system with best available moves to take and thats all it is. It is set of superfast repetitive tasks aimed for one single  purpose only and in every single cycle of tasks it does, lacking one single thing which AI or Intelligence really is...ability of abstract thinking...and thats all there is..just another file processing  machine...if you believe thats an AI, by all means go ahead and be impressed....for sure is nice work done by guys who code it, but its just yet another 'clock cruncher'..

[Steve Elliott]

You don't listen so I won't comment further, just blabbing on about how OLD chess programs work, DeepMind is far more than a chess program or (far more impressively) a GO program, but you don't want to acknowledge go complexity, and how the computer found new ways to play.  It's a GENERAL PURPOSE AI that learns WITHOUT HUMAN INPUT.  It's a step forward in AI.

[3DzForMe]

Roller Coaster Tycoon II, awesome isometric fun.

Glad you brought up the point about running red alert via a VM, I found the game out again and tried installing it on my W7 Box, computer said no... was considering using the game CDs as protection against the birds on my cherry tree... read this thread in the nick of time.

As for AI, the AI in the cars in GRID world series is pretty good, gets my vote. 

[Steve Elliott]

And now I have it on tablet for portable Roller Coaster fun