everyone's doing artificial intelligence wrong

The Something Awful Forums > Discussion > Serious Hardware/Software Crap > YOSPOS > everyone's doing artificial intelligence wrong

«‹›3 »

LP0 ON FIRE: Jan 25, 2006; _{beep boop}

reminds me of all the failed attempts of flying. this is what i see when people try to do AI

https://www.youtube.com/watch?v=gN-ZktmjIfE

# ? May 11, 2015 15:27

Adbot: ADBOT LOVES YOU

# ? May 16, 2024 01:28

LP0 ON FIRE: Jan 25, 2006; _{beep boop}

i look at all these programmers involved in all these overly complicated projects, like getting a robot to half assly walk around and combine other collaborative AI tasks into one to make an "intelligent" robot, but it always results in suckage. and then you have those government funded programs reconstructing the human brain as if it's the ultimate foundation for AI even though humans are capable of the most terrible things and make bad/stupid choices. you really want that to be the future brains of independent robots? lol

# ? May 11, 2015 15:27

LP0 ON FIRE: Jan 25, 2006; _{beep boop}

there will be a smart person one day that will figure out that you need an algorithm to learn by itself and include different senses to understand the world around it. but when someone invents this it will be like playing with fire, and because humans are stupid, we will all be destroyed by it

# ? May 11, 2015 15:30

Carthag Tuek: Oct 15, 2005; Tider skal komme,
tider skal henrulle,
sl�gt skal f�lge sl�gters gang

there are ony two possible outcomes

ether no true AI ever, or it will be made and it will destroy us

# ? May 11, 2015 15:33

Hen: Aug 22, 2006

Everything you need to know about AI you can learn from the ancient board game Go

# ? May 11, 2015 15:34

Necc0: Jun 30, 2005; by exmarx; Broken Cake

so when does artificial intelligence become just 'intelligence' :420:

# ? May 11, 2015 15:37

LP0 ON FIRE: Jan 25, 2006; _{beep boop}

Hen posted:

Everything you need to know about AI you can learn from the ancient board game Go

it's a great example of a board game that a really good AI hasn't been make for it, unlike chess. too bad it's very boring at least to me haha

# ? May 11, 2015 15:38

Necc0: Jun 30, 2005; by exmarx; Broken Cake

also yeah until we can even define what 'intelligence' is how the hell are we going to build one capable of more than playing a board game or finding relationships in numbers

# ? May 11, 2015 15:39

LP0 ON FIRE: Jan 25, 2006; _{beep boop}

Necc0 posted:

so when does artificial intelligence become just 'intelligence'

hmm well it's still made by humans or AI making the intelligence

# ? May 11, 2015 15:39

Sweevo: Nov 8, 2007; i sometimes throw cables away

i mean straight into the bin without spending 10+ years in the box of might-come-in-handy-someday first

im a fucking monster

i took a fuzzy logic class once and the guy teaching it spent the entire time talking about some medical expert system project he'd been involved with in the 80s, and i still dodn't really know what fuzzy logic is beyond "sometimes the answer is in the middle"

# ? May 11, 2015 15:49

GoldenTeeth: Jun 3, 2013; I'm the pensive looking one on the left. That's me. I'm doing serious business in there. Look at me go.

I spent most of my AI class learning about how to solve board games and poo poo

I don't feel like I learnt a whole lot about AI but there isn't a sliding tile puzzle left in the world I can't optimally solve now

gently caress you, professor layton

# ? May 11, 2015 15:57

Nagato: Apr 26, 2011; Why yes my username is the same as an autistic alien who looks like a 9 year old from an anime, why do ask?

actual conversation I had with an AI researcher:

- some of my coworkers came up with a turing test for video games. they made an AI that sometimes attacked irrationally in a fit of rage if it got attacked. most people couldn't tell it from a human player

- well, that's a fun toy, but it doesn't have any practical application

- no it's amazing, i mean this could be a stepping stone to computers that act just like people

- you want your AI to have fits of rage?

- no, but the... uh... principle

- ?????

# ? May 11, 2015 16:01

LP0 ON FIRE: Jan 25, 2006; _{beep boop}

Sweevo posted:

i took a fuzzy logic class once and the guy teaching it spent the entire time talking about some medical expert system project he'd been involved with in the 80s, and i still dodn't really know what fuzzy logic is beyond "sometimes the answer is in the middle"

yeah i don't really get it. fuzzy logic always seems to break down to something that's not fuzzy

Wikipedia posted:

Hard science with IF-THEN rules[edit]
Fuzzy set theory defines fuzzy operators on fuzzy sets. The problem in applying this is that the appropriate fuzzy operator may not be known. For example, a simple temperature regulator that uses a fan might look like this:
code:
IF temperature IS very cold THEN stop fan
IF temperature IS cold THEN turn down fan
IF temperature IS normal THEN maintain fan
IF temperature IS hot THEN speed up fan
There is no "ELSE" � all of the rules are evaluated, because the temperature might be "cold" and "normal" at the same time to different degrees.

"it might be 'cold and normal' at the same time to different degrees"

okay so figure out what degree it is then evaluate?

# ? May 11, 2015 16:01

LP0 ON FIRE: Jan 25, 2006; _{beep boop}

Nagato posted:

actual conversation I had with an AI researcher:

- some of my coworkers came up with a turing test for video games. they made an AI that sometimes attacked irrationally in a fit of rage if it got attacked. most people couldn't tell it from a human player

- well, that's a fun toy, but it doesn't have any practical application

- no it's amazing, i mean this could be a stepping stone to computers that act just like people

- you want your AI to have fits of rage?

- no, but the... uh... principle

- ?????

pretty sure this happened in the 70's on terminals before it became Eliza, and the people on the other end would actually get pissed at the bot trolling them

# ? May 11, 2015 16:03

med school head: Apr 17, 2012

Hen posted:

Everything you need to know about AI you can learn from the ancient board game Go

the only thing to know is that game is for nerd faggots

btw i play Go :smug:

# ? May 11, 2015 16:07

LP0 ON FIRE: Jan 25, 2006; _{beep boop}

gargle chome posted:

the only thing to know is that game is for nerd faggots

btw i play Go

do you play mj? because i know some people that are crazy about that game. kinda interested

# ? May 11, 2015 16:16

jony ive aces: Jun 14, 2012; designer of the lomarf car; Buglord

gargle chome posted:

btw i play Go

how about you "go" out of this thread!!!

# ? May 11, 2015 16:31

Bloody: Mar 3, 2013

this thread is now machine learning chat

pyf machine learning toolkit, library, or environment

# ? May 11, 2015 16:37

LP0 ON FIRE: Jan 25, 2006; _{beep boop}

jony ive aces posted:

how about you "go" out of this thread!!!

omg owned

# ? May 11, 2015 16:42

curse of flubber: Mar 12, 2007; I CAN'T HELP BUT DERAIL THREADS WITH MY VERY PRESENCE

I ALSO HAVE A CLOUD OF DEDICATED IDIOTS FOLLOWING ME SHITTING UP EVERY THREAD I POST IN

IGNORE ME AND ANY DINOSAUR THAT FIGHTS WITH ME BECAUSE WE JUST CAN'T SHUT UP

all we have to do is code a program that will engineer an artificial intelligence to write some AI scripts or whatever

# ? May 11, 2015 17:18

jony ive aces: Jun 14, 2012; designer of the lomarf car; Buglord

Bloody posted:

this thread is now machine learning chat

pyf machine learning toolkit, library, or environment

microsoft azure

# ? May 11, 2015 17:35

Sweevo: Nov 8, 2007; i sometimes throw cables away

i mean straight into the bin without spending 10+ years in the box of might-come-in-handy-someday first

im a fucking monster

all we need to do is write a program that evaluates every possible action. if there are lots of actions then just use a faster computer

# ? May 11, 2015 17:36

med school head: Apr 17, 2012

LP0 ON FIRE posted:

do you play mj? because i know some people that are crazy about that game. kinda interested

i dont acdtually play go i'm normal

# ? May 11, 2015 17:45

fart simpson: Jul 2, 2005; DEATH TO AMERICA

GoldenTeeth posted:

I spent most of my AI class learning about how to solve board games and poo poo

I don't feel like I learnt a whole lot about AI but there isn't a sliding tile puzzle left in the world I can't optimally solve now

gently caress you, professor layton

but thats what ai is though? what did you think you were going to learn about?

# ? May 11, 2015 18:09

Sweevo: Nov 8, 2007; i sometimes throw cables away

i mean straight into the bin without spending 10+ years in the box of might-come-in-handy-someday first

im a fucking monster

how to build my own Cherry 2000 of course!

# ? May 11, 2015 18:19

salisbury shake: Dec 27, 2011

COME BUILD THE VENGEFUL GOD THIS EARTH DESPERATELY NEEDS

# ? May 11, 2015 18:39

LP0 ON FIRE: Jan 25, 2006; _{beep boop}

gargle chome posted:

i dont acdtually play go i'm normal

i was asking if you play mahjong (not the tile matching game)

# ? May 11, 2015 18:48

Symbolic Butt: Mar 22, 2009; (_!_); Buglord

LP0 ON FIRE posted:

it's a great example of a board game that a really good AI hasn't been make for it, unlike chess. too bad it's very boring at least to me haha

this is not really true, I mean if you were talking about it 10 years ago it'd be true, every go playing AI was bad BAAAAAD and anyone who played the game often for like 3 years idk could easily beat it.

but now processors and gpus are much faster than it used to be so a monte carlo brute-force approach is super effective and can beat strong amateur players and some professional players. I guess in like 5 years it'll be able to beat the best human player

# ? May 11, 2015 19:42

Symbolic Butt: Mar 22, 2009; (_!_); Buglord

and yes and I'm a failnerd who plays go sometimes :sigh:

# ? May 11, 2015 19:44

LP0 ON FIRE: Jan 25, 2006; _{beep boop}

hmm ok. still pretty complex

weird to think that checkers was solved like 8 years ago

# ? May 11, 2015 19:48

Space-Pope: Aug 13, 2003; by zen death robot

Symbolic Butt posted:

and yes and I'm a failnerd who plays go sometimes

sup wanna start a go club?

# ? May 11, 2015 20:19

huhwhat: Apr 22, 2010; by sebmojo

Bloody posted:

this thread is now machine learning chat

pyf machine learning toolkit, library, or environment

scikit-learn, because im a bitchmade python babby

also my lab recently did a badly designed matrix game experiment where the algorithm is a better player than quite a few humans since it actually understands the game and doesn't make dumb moves

# ? May 11, 2015 20:28

NoneMoreNegative: Jul 20, 2000; GOTH FASCISTIC
PAIN MASTER

shit wizard dad

salisbury shake posted:

COME BUILD THE VENGEFUL GOD THIS EARTH DESPERATELY NEEDS

settle down there Roko

# ? May 11, 2015 20:30

echinopsis: Apr 13, 2004; by Fluffdaddy

artifical inteligence will desotry japan first

# ? May 11, 2015 20:37

LP0 ON FIRE: Jan 25, 2006; _{beep boop}

then china's AI will start poisoning the earth without any consequence to itself

# ? May 11, 2015 20:46

Salt Fish: Sep 11, 2003; Cybernetic Crumb

Nobody knows what intelligence is in the first place so no duh we can't build it.

# ? May 11, 2015 22:17

coffeetable: Feb 5, 2006; TELL ME AGAIN HOW GREAT BRITAIN WOULD BE IF IT WAS RULED BY THE MERCILESS JACKBOOT OF PRINCE CHARLES

YES I DO TALK TO PLANTS ACTUALLY

'AI' carries a lot of baggage; researchers who entered the field after about 1990 prefer to call it machine learning or computer vision or anything but AI. it's also taboo to make any sort of claim about human brains - that kind of thing caused the winters and the older guys react v badly to anyone that seems to be hyping poo poo

on a philosophical level, the community's also pretty cleanly split between true believers who think 'strong AI' (whatever the gently caress that is) is right around the corner, and the rest who think that our state-of-the-art is laughably primitive

coffeetable fucked around with this message at 22:35 on May 11, 2015

# ? May 11, 2015 22:32

coffeetable: Feb 5, 2006; TELL ME AGAIN HOW GREAT BRITAIN WOULD BE IF IT WAS RULED BY THE MERCILESS JACKBOOT OF PRINCE CHARLES

YES I DO TALK TO PLANTS ACTUALLY

anyway if this is ML-chat: i build ML and CV systems for medical problems. at work i am currently prototyping a structure-from-motion system for [censored], and at home im doing about twenty things which ill post about when any of them go anywhere.

if you are interested in getting into ML, there are two options

black-box approach. install scikit-learn (python or matlab are your options for ML work, dwi), read the tutorials, manually fiddle with parameters until poo poo Seems To Work�. the advantage of this approach is it's fast off the starting blocks. the disadvantage is that unlike most things you deal with as a programmer, ML algs fail silently or are subtly wrong all the time and if you don't know what the insides of the box are then you'll never realise. most ppl who call themselves 'data scientists' fall into this category.
white-box approach. you need to have a basic but very solid understanding of linear algebra and multivariable calculus, and then to work through a book like MLAPP. from there you'll be able to handle the research literature on whatever it is you're interested in. disadvantage of this approach is that it's quite a significant time investment to get going, especially if your maths is out of shape. advantage is you'll actually know what the gently caress you're doing

coffeetable fucked around with this message at 22:54 on May 11, 2015

# ? May 11, 2015 22:46

Bloody: Mar 3, 2013

easy part of ML: the algorithms

hard part of ML: giving the algorithms data they can handle well

# ? May 11, 2015 22:47

Adbot: ADBOT LOVES YOU

# ? May 16, 2024 01:28

Bloody: Mar 3, 2013

feature extraction is in my experience like 80% of the problem

# ? May 11, 2015 22:48

The Something Awful Forums > Discussion > Serious Hardware/Software Crap > YOSPOS > everyone's doing artificial intelligence wrong

«‹›3 »