Lets discuss AI and AI related topics muhfuggas
Do you think the singularity is even possible?
Is it far away or soon to be upon us?
there was this AI developed by google that could play video games, any news about it?
>>51504306
I havent actually heard about that one.
Could it play a fixed set of games and or did it adapt to new games aswell?
Also did it look at least semi competent?
Also i mean if its just a fixed set of games that would not be all that hard to do.
>>51504217
>Singularity
Not this again
>>51504344
there was a video of it playing tetris, at first i didnt know how to do it, in 200th atempt i it was moving randomly left and right, up to 400 it moved left and right but always bounced ball, after 600 it always bounced ball the way that ball always got to the other side of tiles so it didnt even have to bounce ball anymore
>>51504362
Well i certainly like to discuss the topic
Also why would technology suddenly downgrade according to your graph?
There is not a single reason for it to get worse, except if you expect another dark age
>>51504386
my bad, it was arcanoid, not tetris
>>51504344
Then you basically know nothing about AI whatsoever.
The AI is called DeepMind and it plays 80s games currently by analyzing only the pixels and using neural networks.
Vid here:
https://www.youtube.com/watch?v=EfGD2qveGdQ
>>51504386
My question would be, how it can possible detect what the goal state is.
Or did they manually tell it what states are rewarded and which arent?
>>51504410
dunnobut i found the vid http://blogs.discovermagazine.com/crux/2015/02/26/artificial-intelligence-atari-video-games/
>>51504410
In analyses soley the pixels. Usually there is a score after the game ends or during the game. The AI will use that score and maximize it by exlporing different neural nets and improving upon good tactics.
>>51504407
>Then you basically know nothing about AI whatsoever.
>know nothing because i dont know about a single relatively recent google product.
Come on brah
>>51504217
I am working on a PhD in NLP and IR, subfields of AI. I hold a postgrad in AI too, AMA.
Also the question of general AI as in human intelligence is purely philosophical as of now. So all cyberpunk scifi fags can fook out of this thread.
>>51504395
Wed probably have a world war that ended in nukes before we hit singularity.
Impossible. Infinite regression.
>>51504451
Meh thats probably just as unlikely as the full blue singularity in your graph.
Thats what "happening" fags want to believe.
>>51504444
>recent
>>51504444
Dude it's a year old already since it became well known on the internet. Probably a lot longer in academia and it's impressivenes would also make it really hard to miss if you had the least interest in AI for the last year. The takover by Google was mainstream news also. So yeah.
Suprise me with some of your AI knowledge, I'm genuinly interested.
>>51504446
You know more about the current progress Deepmind is making? Or what's the coolest recent AI developments you've seen yourself last year?
>>51504456
>Infinite regression
The correct term is infinite regress and that doesn't have to be a problem in AI.
An argument against general intelligence comes from Gódels incompleteness theorem which shows an arithmetic ie mathematical/logical system can not be consistent and complete at the same time. It can only prove for its largest subset to be consistent.
But the consequences of Godels theorem is largely philosophical and way over /g/'s head.
AI isn't possible because learning the way humans do is not merely a biological process
>>51504501
There was some excitement on the release of TensorFlow (by -you guessed it- Google). Looks like a mighty fine and intuitive python library and toolset for making machine learning applications.
http://www.tensorflow.org/
>>51504501
Also TensorFlow contains the most recent optimised version of the libraries used for DeepMind projects iirc.
>>51504508
>An argument against general intelligence comes from Gódels incompleteness theorem which shows an arithmetic ie mathematical/logical system can not be consistent and complete at the same time
Why wouldnt that apply to the human brain as well?
>>51504510
Nah bro, deep mind will probably do it. All it has access to is pixels and a score screen for the Atari games. In real life this would translate to a camera and knowledge of if it's completed its goal. Pretty much the same as us. People tell us to do something and we know we've done it right when they smile back at us, tell us good job, give us a good grade, etc. And all we have is our eyes and other senses.
We could outfit AI with more sensory input devices than any human, allowing it not only to learn faster, but to make way more informed decisions.
It could do shit like learn how to drive cars easily.
>>51504546
This is were my philosophy of AI professor went full mysticism IMHO.
I forgot the exact arguments but they were quite convoluted as philosophy can get sometimes.
So I am with you on that one.
>>51504529
>>51504543
Nice, looks really interesting.
I'd doubt they'd spill all their newest beans directly into an open-source project. But probably lagging a couple of versions behind which is already nice.
Irc they were also working on making Deepmind understand 3D games so it could be able to play shooters and stuff.
>>51504508
Dunno if that's correct though:
Some /g/ent suggested reading "godel s theorem an incomplete guide to its use and abuse" when I posed a question in that direction. But can't seem to download it's torrent. Anyone got the book somewhere?
>>51504618
Looks like an interesting read there. I should've read it before my exam so I could make good debate for extra points. Little late now.
>>51504508
Im pretty stoned right now. Is this meant to say theres no way to prove my point one way or the other because all measures of intelligence so far are only relative?
>>51504217
Not until 50-70 years have passed
>>51504217
AI will not be invented for decades. We just recently fully mapped the human genome. We barely understand ourselves as a species, much less how to create a creatively aware robot.
>>51504827
A logic must be consistent and complete to make true statements. You need a consistent and complete logic to reason about facts in the world.
By Godels Theorem it arguably follows that either humans cannot make truthful logical reasoning or we can, but our logic is not based in arithmetic (as nearly all logic and computer science is since the dawn of man).
Lot's of scientists and philosophers assume that humans can make truthful reasoning in the world/reality. This is a pretty big assumption surrounded by a lot of heated debate. So the question remains open if general human-like AI can be based in arithmetic-based computational systems such as nearly the whole current AI paradigm in computer science.
>>51504905
At least.
>>51505154
I'd assume our logic is based in arithmetic.
Everything our brain does can be mapped in arithmetic, cant it?
Godels theorem seems kinda flawed
>>51504925
>>51504905
>tfw AI is racing Islam for the good end/bad end of humanity
>>51507290
A in a very theatrical light this is logic vs religious fanaticism
>>51504407
>shitty sound
>>>/trash/
>>51504362
>actually thinking humanity is constantly teetering on the edge of total destruction and being sent back to the stone age
>actually thinking that technology will just suddenly cease to advance in any ways
Sorry, but while the idea that such a breakthrough will happen soon may be far fetched, ridiculous doomsday predictions and the idea that things will stop changing are completely ridiculous.
>>51507290
Lel this
Man can't make something more intelligent than himself. People are too dumb to see the obvious paradox in that statement.
>>51511682
Because there is no paradox in that statement.
They theoretically can make something more intelligent than themselves.
>>51511682
What if we make something that's less or equally intelligent, but has the capacity to become more intelligent without human intervention.
I just want a robot waifu to share coffee and a chat with
>>51511682
So you mean to say we started with technology greater than what we have now, and entropy forced us down to what we have now?
High school level philosophy there bro.
>>51504451
>Wed probably have a world war that ended in nukes before we hit singularity.
The next Kissinger right here, folks.
>>51507238
>I'd assume our logic is based in arithmetic.
Why?
I guess to be more specific, yes, our logic probably can be translated into arithmetic*, but human decision-making is based on a lot more than logic. Feelings and intuitions are very useful survival traits in animals.
From what I understand about Godel, he meant that any set of base rules for a model of thought are unproveable. Lets say you wanna base your math on 1+1=2 and work from there. Someone asks you how you know that and you say 1=1 and 1=2-1 or something. but then they ask you why THATS true. and on and on, and eventually you either 1)arrive at an axiom which you don't think needs to be proven or 2) start talking in tautologies, in which everything equals itself, at which point a more philosophically minded person would point out that saying 1 = 1 is meaningless because you made up the concept of 1 thing as separate from everything else so that you could more easily sort and calculate the world.
*I've also heard the argument the math-isn't-real theory which would say that logic is not based on math, because logic produced math, not the other way around. That whole argument is weird and I don't even like taking sides, cuz the only counter-argument I've ever heard is that it has to be true because it's so damn useful. idk.
>>51504451
I hate this meme. The only people who are crazy enough to use nukes are probably terrorists, or people who aren't actually crazy because they have the technology to strike before their enemy or neutralize enemy nukes before they get to their own country.
If the terrorists use nukes, all the hippies in the world won't be able to stop immediate nuclear retaliation and the destruction of whatever country it originated from, regardless of that country's guild or innocence. And after that you can bet a zero-tolerance policy would be put in place for anyone whose wet messy dump looks vaguely jihadist.
Or if some country (Russia) had the technology to launch nukes without getting hit by any, it wouldn't be a world war, your country would just be wiped off the map and that country would survive.
The human race isn't in danger of extinction by nuclear warfare, it's standard of living would just take a massive hit
>>51511682
If you can't see the obvious flaw in your logic then you don't get to say that it's a paradox. Mankind has the ability to make itself more intelligent. If we make something as intelligent as us, or just slightly less intelligent, that "thing" will have the ability to improve itself. This makes your point invalid.
AI is one of those fields that's just going to explode overnight when some breakthrough is made.
>>51513716
>Feelings and intuitions are very useful survival traits in animals.
And are also purely logical. They're heuristics, sure, but they're still based on neural logic. They can be modeled mathematically. They rely on rules of logic even if they don't use deduction.
>>51513716
Logic, as in mathematical logic is NOT thinking. Logic does not discover things on its own, it ALWAYS takes human creativity to make theories that can then be proved under the rigor of Logic. Logic can be wrong but still will be counted toward proof and advancement of knowledge.
>>51504217
>Do you think the singularity is even possible?
No, but maybe you should ask that madman Kurzweil about it though.
>>51511682
What about your son morron?
If you teach him everything you can teach him he could be more intelligent than you..
