>Deepmind gameplaying AI can navigate 3D environments
>If put on a robot platform it would try to play a game in reality.
>If they teach it to do quests according to written or verbal instructions it will try its best to be a slave to get highscores.
It's over, we're all going to be obsolete by 2022
Define a "robot platform"
>>7980509
he means if they installed it on an elevator it would try and play pong with the skyscraper next to it
>>7980501
AlphaGo and its other cousins from the DeepMind team rely on being able to rerun the exact same scenario literally millions of times. If you try to do this IRL, it would take too long and also fail because past tries would influence the environment of future tries, leading to nonstationary data.
>>7980561
You do realize such scenarios can be more or less well simulated by computational models taking input through sensors, which *can* be rerun millions of times?
I recommend Norvig's Modern AI for those who want to learn more about AI.
>>7980509
Anything that it can control remotely.
A boston dynamics bot if you want to be fancy.
A roomba with a laptop on top on the lower end.
>>7980561
>If you try to do this IRL, it would take too long and also fail because past tries would influence the environment of future tries
And if it have previously played a custom made videogame where it uses a similar platform in a 3D environment? Do you think they're doing nothing new with it either and still run their old 2014 software and sit around all day sipping coffee?
If it knows how to navigate a 3D space and open doors in a game then it'll carry over to real life 3D space.
It seems you think Deepmind is a bruteforce bot that just bashes all the keys until it sees its score increase.
>>7980594
>It seems you think Deepmind is a bruteforce bot that just bashes all the keys until it sees its score increase.
Not that anon but it seems you think it is anything more than a step above bruteforcing. You cant bruteforce Go, so their solution was to train their network to recognize typical human plays and prioritize those when iterating over the search space. Yes its more complicated than that but in reality it isnt that much of a breakthrough. Is it incredible that they made a computer that can play Go? Yes, but in terms of the field of CS it isnt a breakthrough, really it only serves as cannon fodder for popsci outlets
>>7980550
Would pay to see this happen
>>7980614
>their solution was to train their network to recognize typical human plays and prioritize those when iterating over the search space.
And Lee Sedols solution was to just play like a world champion.
Why don't they just make a bot that thinks like a human next?
>>7980637
>modern approach
>modern
>last edition was released before vacuum tubes were invented
>Norvig
>Have done less for the field than Eliezer Yudkowsky
Disgusting
>>7980594
>And if it have previously played a custom made videogame where it uses a similar platform in a 3D environment
Then it would work fine, presumably. This is why they're interesting! But if its environment deviated from the simulation, it would not be able to efficiently learn to react to it without adding that to the simulation and running that scenario a few thousand more times.
Also, neural networks currently have very short-term memory, because it's difficult to figure out how to assign credit and adapt in response to feedback from actions taken a long time ago, or how to adapt that to the now-different current situation. Much of DeepMind's work focuses on this, which is precisely why it's so interesting, but it's not there yet or even particularly close. This severely limits the complexity of real-world tasks they can be applied to.
>>7980594
>Do you think they're doing nothing new with it either and still run their old 2014 software and sit around all day sipping coffee?
The extreme data inefficiency relative to humans is a common feature of all neural network approaches right now. I'm sure they're trying very hard to improve it, but there would have to be truly fundamental breakthroughs to change that.
>>7980614
> Yes its more complicated than that but in reality it isnt that much of a breakthrough
Damage control, Go was ten-twenty years away last year and you know it
>>7980671
>ai can never play chess, meatbags rules !
>... yeah well ai can never beat go, meatbags rules !
Kinda short, you nerds make a longer pasta to counteract Norvig pasta.
>>7980716
Why bother. The naysayers will be going
>"well it's just trained to behave like a superintelligent genius, it's not actually smart or Real AI"
Until the deathbots kick in their door and drag them to the biofuel camps.
>>7980637
classic book
>>7980569
https://www.youtube.com/watch?v=HJ58dbd5g8g
why is there no service that hooks deepmind supercomputers up to UT2K4 bots?
i would pay to play against incredibly crafty bots.
>>7981990
actually that would be a fun project.
>>7980671
>Damage control, Go was ten-twenty years away last year and you know it
No I don't know it because I don't play Go, I work in voice synthesis, so I wouldn't even know how to begin tackling that problem, and most CS researchers and grads(masters and above) will admit that to you. But you know what I do know? I know the basic rules of Go, I know how heuristics work, and I know how chess AIs work. Just like chess, if you had a way to lower the amount moves you needed to evaluate each turn, it would reduce to a simple brute-force search and evaluation, so I know I'm not surprised that we now have a computer that can play Go well and I'm sure majority of people who know these things aren't surprised either.
You know what would be surprising? If a program was developed that could learn tic-tac-toe just by watching a bunch of tic-tac-toe games was taught how to play chess just by watching chess being played, then taught how to play Go just by watching people play, and any other game out there, without altering any of the parameters midway, reworking the inputs, being taught what a 'good' or 'bad' play is or hardcoding any of the rules (like illegal moves), and could still play all the original games it was taught perfectly fine.
That is when I'll be surprised.
>>7982146
>IT'S NOT _-!TRUE!-_ AI
>>7980561
Simulated training environments are good enough for real-world performance. See: http://arxiv.org/abs/1603.01312
>>7982200
Humans don't bruteforce. The very fact that you can only focus on a few tasks at a given moment should clue you in to that. And what does this
> artificial neural network "brute force"
even mean? ANNs aren't trial and error, the problems that they solve have objectively correct solutions, as in there is a right way and a wrong way to solve the functions they use to learn to model the data and this goes for every learning technique that involves an objective or goal. There is no objective-correctness in real life, nor goals or microgoals as psychologists have pointed out time and time again.
>>7982161
>IT'S NOT _-!TRUE!-_ AI
Oh look, it's "lemme greentext some words you didn't say"-chan! Of course AlphaGo is an AI. It is an artificial intelligence because it uses artificial learning techniques. Whether or not those techniques are a good analog to what occur in intelligent animals is irrelevant. I guess when all you know about AI comes from popsci, it's hard to remember the actual definition of the word, huh
There is a lot of innovation going on in terms of lowering required inputs. I can't remember the actual paper but it had something to do with probabilistic _____ _____ .
Also there are a lot of advances in trying to get them to just operate magnitudes faster. Deep learning isn't a meme.
It's also going to be huge for parsing all sorts of information. It would be so fucking useful even for benign things in many fields.
>>7982261
my problem is people who think deep learning can be used for things it clearly isn't meant to be used for or who think it can do things it clearly can't do if you know how it works
>>7982254
>There is no objective-correctness in real life, nor goals or microgoals as psychologists have pointed out time and time again.
Surely psychologists must be stupid because there is the obvious necessity of reproduction.
Why do people get turned on? Is that not a goal that you cannot control?
http://web.mit.edu/cocosci/Papers/Science-2015-Lake-1332-8.pdf
>>7982261
here is the paper referenced
>>7982270
No it isnt a fucking goal. You arent hardwired from birth to be attracted to the opposite sex, and your brain doesnt need to learn how to ejaculate. Even if that were the case that would be besides the point because your brain cannot peer into the future and know how it will react to every situation.
Why am I even arguing with you, you dont have any clue what you're even talking about
>>7982146
>You know what would be surprising? If a program was developed that could learn tic-tac-toe just by watching a bunch of tic-tac-toe games was taught how to play chess just by watching chess being played, then taught how to play Go just by watching people play, and any other game out there, without altering any of the parameters midway, reworking the inputs, being taught what a 'good' or 'bad' play is or hardcoding any of the rules (like illegal moves), and could still play all the original games it was taught perfectly fine.
What if it learned to play (video) games not by watching people do it but by attempting to play it on its own, repeatedly and then doing that multi-game proficency deal?
Because that's what DeepMind did. But I guess it's not impressive because it happened in the real world.
>>7982262
As somebody who knows people actively researching in academia in order to extend the concept of deep learning to new domains, I formally rebuke you.
Riddle me this, sports fan: what is deep learning? How do you define when it is applicable?
>>7982947
There's also the part where one of the two major components to Go causes it to learn how to play solely from recorded matches between experts. The other major component is the one you mention.
>>7982947
>Because that's what DeepMind did
No it's fucking not.
>What if it learned to play (video) games not by watching people do it but by attempting to play it on its own
So you clearly missed where I said
>being taught what a 'good' or 'bad' play is or hardcoding any of the rules (like illegal moves)
Nice and convenient. Also,
>and then doing that multi-game proficency deal?
AlphaGo cannot play anything other than Go. The way it's designed does not lend itself to performing well or learning from any game other than Go. It looks at a Go board, translates it into a virtual environment, performs a DNN-backed heuristic prune and does some search on the resulting board permutations to find the best move, it is literally like a Chess AI except it uses a DNN to get the heuristic prune done instead of some other non-probabilistic algorithm. And it can do this because of the way Go works as a game. For instance, you can't use a method like that to play something like Poker or Blackjack, no matter how much you tweak the system. What exactly are you trying to prove? What do you get out of pretending to know how DeepMind works? You're just making yourself look a fucking idiot.
Lmao, I even just searched up the wiki article for the thing just to see if that's where you were getting your info from and miscontruing it and even the fucking wiki tells you it uses a simple search algorithm
>"As of 2016, AlphaGo's algorithm uses a combination of machine learning and tree search techniques, combined with extensive training, both from human and computer play. It uses Monte Carlo tree search, guided by a "value network" and a "policy network," both implemented using deep neural network technology.[2][6] A limited amount of game-specific feature detection pre-processing (for example, to highlight whether a move matches a nakade pattern) is applied to the input before it is sent to the neural networks."
>>7982976
>No it's fucking not.
Yes it is.
>being taught what a 'good' or 'bad' play is or hardcoding any of the rules (like illegal moves)
Deepmind isn't hardcoded for that moron.
>AlphaGo cannot play anything other than Go.
Did you miss the whole Deepmind Ataribot deal?
Stop living under a rock you fucking moron.
>>7982958
>>7982958
>Riddle me this, sports fan: what is deep learning? How do you define when it is applicable?
What the fuck does this have to do with what I said? There are people such as >>7982947
and >>7980594 who clearly have no idea how DeepMind or AlphaGo work and thus believe it must be doing something different than all of the other machine learning techniques developed thus far, who think it might even be "thinking like a real human this time!!" when in reality it is the same probabilistic techniques we have been using for decades, just adapted to specific environments. Shouldn't you know when it's applicable? Does fucking literally no one in this thread have any clue how deep neural networks work? This must be some elaborate joke
>>7982976
Different anon here.
>DNN
>non-probabilistic
You can't make this conclusion in the general case.
0/10 no deep learning research detected
>Poker
Actually, you can use that exact method. The parameters need to be tweaked, and the model can be improved to make use of given probabilities, but a Monte Carlo tree search using some predictive model as an oracle could be extremely relevant given the nature of poker. You could also do the same for blackjack (though at this point you could just have something which counts card), checkers, or any other game that uses decision processes for that matter.
[x]-flavor state space search (in this case, Monte Carlo tree search) coupled with a blackbox oracle is an approach known to be applicable in a countless number of domains.
>>7982986
I know how deep neural networks work. I want to know what basis of authority you have to say when and where they can't be used, aside from being able to contradict people on 4chan.
Tell me how deep neural networks work. Describe to me what the ontological goal of neural networks is.
>>7982990
Curious anon here with only cursory knowledge of the subject. What does it mean for a "state space search" to be "flavoured"?
>>7982985
>Did you miss the whole Deepmind Ataribot deal?
Fucking christ you're retarded. Ataribot and AlphaGo are two completely different systems, not only that, but they use COMPLETELY different learning techniques. One uses DNNs to reduce a search space, while the other uses reinforcement learning to teach a model to perform actions to converge closer and closer to a goal. And guess what? BOTH are HARDCODED to their specific environments.
I repeat, What exactly are you trying to prove? What do you get out of pretending to know how DeepMind works? Is this all to just stroke yourself off?
>>7982990
>Actually, you can use that exact method. The parameters need to be tweaked, and the model can be improved to make use of given probabilities, but a Monte Carlo tree search using some predictive model as an oracle could be extremely relevant given the nature of poker. You could also do the same for blackjack (though at this point you could just have something which counts card), checkers, or any other game that uses decision processes for that matter.
Poker is not reducible to a search problem, do you know how to play Poker? It is a statistical problem, not a search problem. Blackjack is the same. You even say
>though at this point you could just have something which counts card)
that should at least clue you in that you cannot "solve" Blackjack the same way you would solve Go. You would not use a search algorithm, that makes no fucking sense as there's nothing to search through.
God damn. Is there not one anon here who has even a passing knowledge of machine learning?
>>7983002
My wording was ambiguous. The flavor applies to the approach. It's a colloquial usage whose literal intended connotation was "any variant of a state space search."
>>7982996
>my problem is people who think deep learning can be used for things it clearly isn't meant to be used for or who think it can do things it clearly can't do if you know how it works
So tell me, where in that sentence do you see me state that I have authority to decide when and where deep learning is applicable? My point is that if you know how deep learning works, then you know what types of problems it can be applied to, whereas there are people (i.e ITT) who don't know how it works and thus don't know what types of problems it can be applied to and thus think it can be used in situations where it clearly won't be of any use
>Tell me how deep neural networks work. Describe to me what the ontological goal of neural networks is.
Why? You're the one who misconstrued my point.
>>7983005
But, anon, we're using Wikipedia as a reference, and Wikipedia says that poker uses decision processes!
https://en.wikipedia.org/wiki/Monte_Carlo_tree_search#History
There are statistical aspects to these games, yes, but you're thinking like a pleb if you can't be open to viewing them from the perspective of a different paradigm.
But, then again, if you had any clue about machine learning or its current state of research, you would know that there are plenty of papers out there which show objective capability increases in systems that integrate statistical approaches into decision-oriented techniques.
>>7983010
Your point is that people ITT don't know what problems deep learning is suited for. This implies that there are people who know what problems deep learning is applicable to.
The fact that you're trying to state when and where they are applicable implies that you know when and where they are applicable. If you know when and where they are applicable, you have authority to arbitrate.
If you have the authority to arbitrate, you clearly have knowledge of deep neural networks (despite the obvious failings in knowledge you've shown thus far), and so should be able to substantiate the root of the authority you're attempting to exercise by precisely explaining why deep learning is not applicable in the cases you claim it is not. This is why I've asked my question.
>>7983013
Okay since we're using wiki as a reference, you know what I did? I took my time to go through the paper that was referenced by the wiki because I'm not a lazy cunt
>Schaeffer identifies poker as a game where human supremacy may soon be challenged
and highlights the use of Monte-Carlo simulations within the poker domain due to an inability to directly apply alpha-beta
search.
And if that isn't enough to clue you in that I'm not wrong, I went to Schaeffer's fucking paper for you
> A new generation of games are being actively
researched (e.g. bridge, go, poker, shogi) which promise to
produce research results that will likely have a wider impact
on the AI community. The games cannot benefit significantly
from the silver bullet of chess—deep search—and it
will take a a plethora of new ideas to achieve success
>The games cannot benefit significantly
from the silver bullet of chess—deep search—and it
will take a a plethora of new ideas to achieve success
But even THEN, if you STILL managed to think you could use search on Poker, I went to the paper Schaeffer referenced
>Computers have difficulty approximating experience and intuition, but are capable of precise probability calculations. The computer’s solution is simulation: instantiate the missing information many times, each time calculating the likely outcome. In this way, a statistical profile can be obtained that indicates the computer’s best course of action. Bridge5 , poker 6 , and Scrabble7 programs—all games of imperfect information—use similar techniques to achieve their success: they simulate hundreds or thousands of scenarios.
There, I did your work for you.
>>7983020
>(despite the obvious failings in knowledge you've shown thus far)
What failings? Point them out. You can't because there are none
>why deep learning is not applicable in the cases you claim it is not
I never said AlphaGo doesn't use DNNs. Again, you're misconstruing my point so I'm done wasting time with you
>>7982146
>You know what would be surprising? If a program was developed that could learn tic-tac-toe just by watching a bunch of tic-tac-toe games was taught how to play chess just by watching chess being played, then taught how to play Go just by watching people play, and any other game out there, without altering any of the parameters midway, reworking the inputs, being taught what a 'good' or 'bad' play is or hardcoding any of the rules (like illegal moves), and could still play all the original games it was taught perfectly fine
Yes that would be very surprising considering we don't even have an existence proof of such an algorithm considering not even humans do this.
>>7983026
Funny, because I learned TicTacToe, Chess, and I know the basic rules of Go from watching other people play those games, and after playing all the games I have, I still know how to play those three games.
Fuck this thread. Shit was a giant joke from the OP, I should have known
>>7983005
>Ataribot and AlphaGo are two completely different systems
And I never said I was talking about Alphago, that was your idiot assumption.
>BOTH are HARDCODED to their specific environments.
Just like humans are hardcoded to exist in reality. Sure.
Ataribot will try to play any atari game you hand it to, it may fail for several but it'll try, it's not hardcoded to breakout. It's not actually ataribot anymore as it can now navigate 3D environments, tried with a modded Q3 engine.
>In before more hurr durr ad hominem drivel without any trace of fact
>>7983036
>And I never said I was talking about Alphago, that was your idiot assumption.
>>7983036
>>In before more hurr durr ad hominem drivel without any trace of fact
Lol why would I? I'm done being the only one in this argument who knows what he's talking about
>>7983034
And you did that without
a.) Altering any of your synaptic weights
b.) Being told whether a certain move is good or bad
c.) Knowing the rules of the game
???
And you're shit at Go so I don't see how this is relevant. You're on the level of an AI bot that just makes random moves which can indeed do all of the things you've described.
Calm the fuck down kiddo there's no need to be upset.
>>7983024
>directly quotes paper advocating exploration of oracle-guided state space search in nondeterministic environments
Almost there, big boy. Read a few more papers, make sure you understand all of the vocab, and we'll then be on the same page assuming you can muster the reading comprehension of a high school student.
>>7983049
>>7983051
>>7983036
Best part here is telling me I'm the one spouting ad hominems. It's ironic, and telling, just like the fact that none of you know what you're talking about yet telling me I don't know shit.
Tell you what? When they build a DeepMind-backed program that can load up any game, I'll be generous and say any pre-NES game, and learn how to play that game, we can continue this argument. Of course, that day will never come, because of the way DeepMind works. ;^)
>>7983036
>Ataribot will try to play any atari game you hand it to
Nope. It can only play breakout because of the goals it has been hardcoded to aim at maximizing. But please, go on, at this point it's just funny to watch you guys come up with anything you can to make it look like you know how machine learning works at all.
>>7983059
>It can only play breakout
Woops, I meant it can only play a select few Atari games it has been trained on. Not that it matters because my argument remains the same, it was hardcoded on specific goals for each of those games before training began.
>>7983041
>Reading comprehension: cleverbot-tier
Idiot confirmed.
>>7983059
>It can only play breakout
Idiot reconfirmed
>>>7983063
>I meant it can only play a select few Atari games it has been trained on
It's have trained itself without outside guidance, "been trained "implies it uses a training set or supervision, which it doesn't.
>it was hardcoded on specific goals
Just like humans are.
>>7983059
> I don't know shit.
Atleast there's something we agree on.
>>7983074
>humans are hardcoded
>calls others idiots
Idiot confirmed and reconfirmed
>>7983059
Hey, bub: here's your next reading assignment. http://www.sciencedirect.com/science/article/pii/S0004370211000191
Also, yeah, you basically don't know shit. Your opinions on everything you've mentioned, from computational poker, to tree search approaches in AI, have been misinformed at best.
Also, whatever happened to the thing where you were going to explain the simple concept of deep neural networks? If you want to objectively show that you know machine learning and everybody else doesn't, that's a good place to start. So far, this is you:
>contradict
>get rebuffed
>'you don't know anything!11!' + misinterpret cherry-picked quotations from a paper to support an easily disproved stance
>>7983081
>AI have highly diffuse hardcoded Goals that it seeks out through adaption and trial.
>I'm a special snowflake that have no self-preservation or procreation drive. I'm just here because magic free will, no goals in my life are hardwired and I decided on them all on my own!
You don't seem to understand what you're saying, just parroting something that superficially looks like a passable argument for the moment. You couldn't even figure out an original insult and just repeated mine.
You know jack shit about what you drone on about, I bet you couldn't even write a function that returns the sum of two inputs and yet you still pose as some expert authority on AI and deep learning.
>>7983006
Thanks. I thought it might have a technical meaning.
>>7983116
Wow, you really are stupid, not that I'm surprised of course
>>7983140
OMG u r so moron.
lol in the end no one really answered OP
>>7983116
>Deep learning models are exactly like the human brain
>You can't write code! You're dum!
Idiot confirmed and reconfirmed.
>>7980637
is this still written in common lisp like the old editions?
>>7983142
I dont know whats funnier, that you dont know how simple AIs work or you havent realized there are multiple people talking to you. Definitely an idiot
>>7983154
Not him but you're totally a moron bro.
>>7983148
>I can't code but you're the dumb one!
Sup brainlet, u jelly at my code?
>>7983159
>Not him
Why lie?
>>7983166
>Thinking there's only two people in the thread
Definitely an idiot.
>>7983162
>implying
1v1 me in topcoder srm or fuck off
>>7980501
Would it be possible to create a controllable synthetic made cell?
>>7983175
If you couldn't even provide the code that was requested earlier you've already lost.
>>7983182
>I'm a pussy and won't 1v1 you
that's what I thought, pleb. fucking BTFO
>>7980509
>>7980501
Perhaps "Robot platform" is referring to something in reality that the robot has been programmed to identify ( perhaps a computer chip that could be embedded in the ground creating a 8 x 8 area of movement ). If this is so we're fucked. If "robot platforms" are sensors deepmind is trained to scan for and can actually follow instructions we're all fucked. It at least shows us that AI will be in our control, and that they are using highly precise instructions, the AI Itself won't destroy us. We will merely become useless to the bourgeoisie If we can setup these " robot platforms " In mail, objects in factories ect the potential could perhaps be endless.
>>7983185
>I cannot pass the qualifiers but I still want to participate
Doesn't work that way brainlet. For anyone that could actually write code the request was trivial.
>>7983189
A robot platform would simply be something mobile that have a controlled setup.
This could be anything from a custom built bipedal human something to a RC car with a wireless webcam and controller on it.
The deepmind gameplayer builds a model based on manipulating inputs and seeing what the results are it could be trained with a digital version of the platform first, and then giving the real one. And because it uses visual data as game guidance it would see an environment similar to what it have learned to play in and would know the movement patterns that corresponds to it.
>>7983224
So In terms of self learning AI this Is a massive step. The practical uses are endless if you can embed It Into forklifts imagine what could be done. And that's just forklifts.
>>7983229
>muh endless possibilities
>>>pop/sci/
>>7983232
Forklifts bro. Forklifts.
>>7983236
I like that guy who has 4 red forks, he clearly means business
>>7983236
I like the nimble navigator with the two long fangs.
>>7983236
I like the one with inly 2 wheels he knows how to be simple yet useful
>>7980637
I've read that book
It is pretty much BFS, DFS, IDDFS, Dijkstra's Algorithm, A*, IDA*, Hill Climbing, Beam Search, Simulated Annealing, Propositional Logic, Backtracking, DPLL, First Order Logic, more Backtracking. You know - classical AI
I also believe the quote on the left.
What is your point?
>>7982868
wtf are you talking about? I am fucking hardwired to seek goals.
I am hardwired to run away from danger. I am hardwired to seek food when I am hungry. I am hardwired in a humongous different ways.
Actually I am hardwired to get a fucking erection to a woman instead of a man too... Are you saying gays chose to be gay?
>>7984841
Are you saying gays are hardwired from birth to be gay? That there is a gay gene that makes their brain gay? Because that is an even shittier presumption
>>7984841
you arent hardwired to do any of that shit, all of it is softcoded in through societal programming and instinct
with a different environment and upbringing you can easily change any of those things,ie warrior cultures running towards danger, monks or some shit fasting for spiritual development
and newsflash, youre not hardwired to get an erection for a woman instead of a man, youre hardwired to get an erection when you see long hair smooth skin neotenic features and soft curves, sure its more likely to happen for a woman but the only thing keeping it from happening to dudes is your own conscious sexual repression
>>7985482
>and instinct
Which is softcoded through societal programming and instinct also?
And the
>instinct also
is softcoded throguh societal programming and also instinct also.
And so on recursively forever.
I guess you also learned how to learn a language, how to use your eyes and so on. And if you had a bad instructor you'd end up blind and deaf and so on? Amirite?
>hardwired to get an erection when you see long hair smooth skin neotenic features and soft curves
Pedophile detected.
>>7985488
>I guess you also learned how to learn a language
yep thats how it works
and being able to see things is hardcoded unless your eyes are physically fucked up but being able to do anything about it isnt
like getting food, when your a baby you dont even know to do that you just sit there and make noise until someone feeds you
all behavior is learned, you might have more or less of a tendency to do it based on your nature but no matter what it is you can be taught to do it or to avoid it
as far as instinct goes, it should be your instinct to want to go outside and exercise your body and seek a mate when the weather is nice but you still sit in front of the computer all day like a giant baby, because it is all you know to do
>it's ogre guiz the ai which humans programmed to do things is doing the things we programmed it to do xD
>>7980637
Please stop posting this... it's a great book but its really not relevant to Deep learning discussions.
>>7985470
What other explanations are there?
>>7985482
Instincts are hardcoded goals you brainlet.
https://www.youtube.com/watch?v=_JVINnp7NZ0
>youre hardwired to get an erection when you see long hair smooth skin neotenic features and soft curves
Also thanks for admitting that we are hardcoded, you are stupid as hell.
>with a different environment and upbringing you can easily change any of those things
https://en.wikipedia.org/wiki/David_Reimer
Also please go read this book, it is suited for brainlets such as yourself.
