Artificial Intelligence

>be me, Kasparov
>number one chess player
>no one can beat me, not even AI
>IBM creates Deep Blue
>1997, I am world champion
>Deep Blue challenge match
>is a supercomputer
>uses minimax search with alpha beta pruning
>has iterative deepening search with heuristic function
>beats me at my own game
MFW

Attached: MFW.png (840x554, 1.08M)

Other urls found in this thread:

youtube.com/watch?v=Z-HL5nppBnM
youtu.be/lb3_eRNoH_w
en.m.wikipedia.org/wiki/New_Chronology_(Fomenko)
twitter.com/NSFWRedditImage

so what? is technological advancement not a good thing?

Mfw it was Bobby Fischer controlling the computer

Making an unbeatable chess program isn't really true AI though. You would have to make a system that wants to beat you at chess and can decide on its own whether it wants to play. And then it would have to be able to learn and master other games at will, based off what information its gathered through other experiences.

today AMD FX6100 dont give any chance to any human

Attached: ESP8266OLED.jpg (1000x1000, 153K)

have nothing where human better tham mashine now

>Game has a large, but finite number of moves and outcomes.
>Computers handle these exact parameters on a regular basis.
>Time, study, and programing later...
>"Waah! How did this happen!?"
This is not A.I. It is basic (though to a human, tedious) problem solving.

You're next, Go. I can already taste your weeb-flavored tears.

Go has already been conquered, where have you been the last few years?

Chess is basically tic-tac-toe. If you can know and memorize all possible combinations of the board (lol good luck) you'll be the best *possible* chess player.

>You're next, Go.
youtube.com/watch?v=Z-HL5nppBnM

>fucks your wife

This has been proven.

They did try to run it as an AI, but when it lost the first couple of matches they gave control to Bobby using the computer just as an extra aid.

unfortunately there are more possible chessboard combinations than there are atoms in the universe

Under a rock, apparently.

AI =/= superintelligent humanlike machine

an AI doesnt have to "decide if it wants to play" at all. Did you watch a lot of sci-fi movies lately?

The definition of AI keeps shifting.
Initially a vending machine that could calculate the correct change was considered "AI".

It's basically "AI" until someone figures out how to make it, at which point it becomes an "algorithm"

Not if you only count moves that would happen in a real match.

The vast majority of moves are too dumb to even consider.

There were allegations of human interference with Deep Blue, Kasparov asked IBM to release the logs, they refused.

Until IBM comes clear with this there will remain a big shadow of doubt over Deep Blue.

>Chess is basically tic-tac-toe.
Spoke the brainlet.

my friend showed me a game called dota where an ai called open ai beat supposedly top class player, he said that ai is same as alpha zero, it learns from patterns instead of algorithms

it is though and that's how they play it which is why it gets boring after a while unless theyre autismos that would otherwise have 20000 hours in dota in another life

Bingo.

nice

>based off what information its gathered through other experiences.
Already happening

youtu.be/lb3_eRNoH_w

Pretty damn amazing honestly. BTW the guys in thos thread saying how the game it's so easy for a computer because "there's a finite ammount of moves (or sensible moves) and it's like tic tac toe"... lol. You don't know what you're talking about, just stop.

algos are patterns :^)

Chess isn't even hard

Krapsparow prearranged all his matches with Krapov including the IBM ones.

In the sense that it has a "solution", yes. it is exactly like tic-tac-toe

t. bobby

T. 1300 elo player.

Who cares? You're competing against humans, not computers, Those AI tournaments are just publicity stunts.

That's literally not how AI is defined by anyone.

Even positions that make sense are more than any computer can fathom. Bruteforcing chess is a horrible idea

You do know he was harassed by the IBM team before each match, RIGHT?

Attached: 39548F18-DBCC-402F-8D32-E873E8279320.jpg (640x408, 37K)

Art

>be me, Kasparov
>lose to AI
>suddenly become respected AI figure, despite not knowing anything about AI
>cash out on fame

It's called artificial intelligence for a reason. It's not real intelligence, it's artificial.

It uses machine learning to adapt to the game, that is what AI means; using machine learning to "learn" new processes
Please reffrain from posting again outside of Sup Forums

Really? How'd they bully him?

HURR DURR EVEREYTHING THAT HAS A "SOLUTION" (HAHA I PUD IT IN QUOTES LOL) IS THE SAYM AS TIC TAC TOW XDDDDDDF

Attached: CE3DFD9A-13EA-4278-8D2C-F1DBA40D5649.png (348x349, 5K)

>t. chess fag who fell for the chess meme

They already have a bot who learned how to play DOTA2 and beat one of the world's best players already.

Nope. Bobby was brought to supervise the moves and gibe feedback to the programmers

There was a good technical thread about AI yesterday:

well the guy believes the jews faked 1000 years of history, no wonder he lost

kaspersky? no wonder he's so keen on america

Yeah, I'm satirizing but there's really a theory. Silly slavs
en.m.wikipedia.org/wiki/New_Chronology_(Fomenko)

Art dont give any money so it not even try to computeresy it
shure compuer can do it better

Attached: _____.jpg (1300x956, 199K)

:O

math
also making computers

>learns from patterns instead of algorithms
an algorithm is literally just a way of solving a kind of problem, everything a computer does is algorithms. in order for a computer to 'learn' anything it needs algorithms

brb gonna bully my way to grand master

Not one of those guys, but can you enlighten us?

No it's not, haven't you read Kaczynski?

Not for long

Just wondering, how is a modern average PC compared to a '97 supercomputer? Could a top of the line i7 have beaten Kasparov too?

Machine learning and AI are not at all synonims. Machine learning is a subset of AI, the same as searching. Deep Blue used Minimax which is s search algorithm, it's still AI, but not ML.

Fucking based

Mobile phones today play at a level far beyond the greatest player in the world.

he made a stupid mistake and machine has all the master moves saved.
It isn't AI

Install the play Magnus app. At about 16 years old almost no human can beat it. Magnus himself posted a video of him playing it at age level 14 and losing

An AI in 1v1 dota beat a human. Completely raped him. May have been different if another hero was chosen.

I don't think it would be that hard to work out positions that make sense. I don't think we're including just "bad moves" but literally virtually impossible such as all black being on the white side.

All this Dota stuff makes me curious on how Ai would perform in other games. What of tactical games like Disgaea, Fire Emblem, etc.?

Attached: Most Curious - Eru.jpg (637x685, 77K)

Grime on Numberphile mentioned an estimate of not shit moves, but I can't remember the exact number. I think it was at least a googol.

No, of course not.

I get the feeling you could brute force specific levels in tactical games pretty quickly with a machine learning algorithm trying to play through the game enough times. There's only so many options available and it should recognize general patterns on approaching enemies.

There's a couple Smash Bros Melee AI bots out there that are cool to see as well. One guy streams it training on itself sometimes

They gave him cold tendies

Attached: anger.jpg (269x211, 8K)

you should actually take a look at the dota stuff then and how the ai actually won and the restrictions that its opponents were placed under. The match was 1 v 1 which isn't how dota is normally (5 v 5) played but professionals have good mechanics anyway so they will generally be good at 1 v 1. The ai was beaten by someone using an incredibly cheesy move that would never work against someone that was slightly skilled and the ai couldn't react to it at all. Current AI does really well in small tasks but when the number of unknowns increases then it shits the bed hard. If an ai had to learn a game from scratch with no user inputs, then if that game has fog of war then it will have a very hard time.
the company working on the dota bot want to make it play 5 v 5 and google want theirs to play starcraft II. Both ai's are struggling. they can play the minigames at an insane level but they have no concept of strategy or teamwork.

Someone who wears underwear under their kilt isn't a true Scotsman though.

Yes and that's fine. I'd love to see something like the Wars series being picked up and see what the AI does to maximize rank or just straight up do low turn count.

Will do. It's crazy difficult to have them work them all together, I don't know where I would start if I were given the task. The 1v1 was optimized for that, simply throwing it into a team scenario is not good at all.

>MFW some faggots spent years perfecting the world's strongest chess engine that even wipes the floor with Magnus Carlsen
>They spent months writing complex algorithms that analyze the board up and down
>They hardcode pages upon pages of opening and endgame theory into it
>Yet Alphago plays with itself for a few days from scratch using relatively simple NNs and absolutely demolishes stockfish at its own game

Attached: BLOODBLOODBLOOD.jpg (164x162, 16K)

Better processing power

>Yet Alphago plays with itself for a few days from scratch using relatively simple NNs and absolutely demolishes stockfish at its own game
No. It did not demolish Stockfish at its own game. And it probably isn't your fault you didn't know this, because otherwise click-bait titles wouldn't have been as effective.

The game was played at a fixed time of 1 minute per move, so this eliminated one of Stockfish's advantages, it's time management heuristics. Stockfish also had to play with more threads, but was only given 1 GB of cache, so the hash tables too small for it. This was requested by AlphaGo so they could be "equal", but actually, it was to AlphaGo's advantage.

Also, AlphaGo was using custom hardware designed by Google -- 4 Tensor Processing Units vs 64 i9 cores for Stockfish. An AlphaGo TPU is capable of processing 180 TERA flops, compared to a Core i9 CPU able to work at 100 GIGA flops. Tera vs. Giga.

So ... yeah.

.t stockfish dev

>lul why did the cavemen not just use lighters
are you mentally handicapped? or maybe that was supposed to be an insult? technological progress happens in stages. each advancement is made possible because of previous iterations. not to take into account how the match actually played out.
learn to think before making such stupid comments

.t second stockfish dev

Yes, it is, but MUH FEELS THO! Muh pure state of nature, muh mud brick house by the river, muh simplicity!

Wow you are so wrong I cant even

Used to say language, but your use of the english one shows thats not true for everybody.

Some players actually menaged to beat the bot without any cheesy strategy back before it went public.

They just represent different ways of attacking the problem.
One is viable on ordinary PCs and the other is custom designed ML hardware that no one can buy.

You can rent TPU power for 6.50 an hour per chip

You can, but it's expensive. Plus I don't believe the second gen chips can be used yet? I haven't kept up.

Man may not be replaced

Attached: 1j185m.jpg (800x340, 45K)

We'll loop back to simplicity with technology so advanced it disappears into the living environment. Remember the phone from Her?

Alternatively
>algorithm + lots of data = machine learning
>inb4 we don't know how it works exactly but it gives good result so we use it anyway

didn't kaspersky have a rematch and win against Deep Blue?

>Initially a vending machine that could calculate the correct change was considered "AI".
What? Isn't that just basic programing?

That's not really applicable in things like video games though where computers outclass humans in reaction time and micromanagement.
A computer will easily beat any human at any fps game by having perfect accuracy

This.

I didn't had a better word but as my friend described, he said even in a game of chess that ai doesn't use mathematical approach but uses patterns in a way like "if that happens than what will happen" well I don't really know much about that but I thought you guys will like it

>Yes and that's fine. I'd love to see something like the Wars series being picked up and see what the AI does to maximize rank or just straight up do low turn count.
You remind me that is is a stupid idea I once had. I do think I'll be able to create something decent given enough time. I'll add it to the list. Maybe I'll actually do it one day.

You know how powerful cheaters with aimbot and shit appear to be at every game? Like they can 1v5 or 1v10 ? Now give that cold precision to a machine that will shoot you in the first nano second it sees you. Just OP

Esp32 is better

An artificial intelligence should be aware of its own perceptions, be able to question their validity, and furthermore understand that these perceptions are its own, and therefore question itself. That rules out alot of animals and insects btw.

the 1997 Intel ASCI Red/9152 could do 1338 GFLOPS
a GeForce GTX 480 can do 1345 GFLOPS

double price for uselees features

Citation needed

if you can't make a move in a minute you're worthless