s/fur
I feel like doing an AI experiment, but how do neural nets learn?
Has anyone gotten over the hurdle of always having to try thousands mutations until something works, or are there AIs that after a bunch of lucky mutations are able to improve on the fly?
Whenever I see a picture of a neural net, I always see layers.
I suppose you can't take pictures of more sophisticated AIs, but the layers kind of imply that no one is making fully self-modifying AIs.
Actually, how do human brains learn?
Neurons aren't connected at random, that's for sure. But they multiply and connect. What process governs that?
>>742463194
Learn about Backpropagation
>>742463194
Off to work so goodluck op
>>742463275
Also our learning comes from strong emotions and sleep. You learn more studying at night.
Gentlemen.
moar like this
I'm single again!
>>742463747
I align my long and hard 0day exploit with your backdoor.
Penetration is so hard and fast, it reaches your NT AUTHORITY\SYSTEM int one motion.
I don't stop there. Using my row hammer technique I constantly assault your io streams.
You are close... I can feel it. your interrupts are off the charts.
In last moment I logout and bsod all over your lagging gui.
And as finishing touch I login in your violated winlogon and do a taskkill /PID 4 /T
>>742464351
Wouldn't it be better to use /F?
>>742464417
no need.
>>742463747
corn?
Any Roblox players here?
99.999% of the games are complete shit but it's worth it for the rest.
>>742464351
Amazing. Bravo.
>>742465183
Gotta love those colors
How do I get better at loli voice acting?
>>742465369
squish your balls?
>>742465505
I'm pretty sure that's just a comedy cliche.
so, what are you guys listening to?
>>742465657
Have you tried?
Post the deepest song you can sing
Post the highest song you can sing
https://www.youtube.com/watch?v=BEm0AjTbsac
https://www.youtube.com/watch?v=5bHimOJb-Xw
>>742466149
Slightly rapey stuff is just
unf~
post more tickling
>>742466312
Me too!
I'll see if I got anymore!
>>742466422
Not furry, but this is pretty great.
>>742463194
Neural n-networks learn by modifying the weights their internal functions use t-to decide output b-based on expected error vs actual error.
Think of it like a coach telling you when you messed up a jump shot vs when you did it correctly; you modify your posture based on how "right" you got it until your start doing it correctly.
D-depending on what the internal algorithm is for each neuron (LSTM, GRU, etc) and the way errors are propagated up (forward, backward, feed through, and many others), you g-get different error rates, different l-learning rates, and varying degrees of overfitting.
R-recommended reading along w-with simple to use API for training your own neural networks (of type RNN with LSTM): http://karpathy.github.io/2015/05/21/rnn-effectiveness/
>>742466562
That's still just mutation though.
Mutation is always necessary of course, but the more control you give an AI over its own structure, the higher the potential rate of adaptation is.
The cost though is that it takes longer for the AI to reach a point where it's not goofing around and being completely useless.
>>742466749
*blinks* It's n-not mutation dear; mutation is random changes, such as are in genetic algorithms or simulated annealing. T-there need not be any randomness involved in t-the training of a neural network.
Take an LSTM for example; it keeps errors around for a fixed amount of time, to allow back propagation to get more of a chance to fix errors. There is no randomness in this process; the neuron simply carries more information and uses it to adjust itself as more and more data flows through.
I think perhaps you are making a simple categorical error <4
Euro friendly fur thread, count me in.
>>742470305
>dress
>beer
>no underwear
My dreams have collapsed into one picture.
>>742471421
>>742466019
I used to sing more than I did, and trained my vocal fry to be pretty powerful down to about G1, and I have a natural whistle register which is very rare for men. I'm not very good though, and it's been a while since I've practiced, but it's just over 6 octaves: D1-E7.
Peter Steele is much bassier than me, but I can sing his low stuff decently.
https://www.youtube.com/watch?v=RLfa5V0uF-Q
I don't know of any songs with an E7, so here's Tom Araya hitting a whistle C7 for comparison.
https://youtu.be/sYeoI-kRjjU?t=22s
>>742472504
Well shit, you blow even my already large vocal range out the water.
Still, I can easily speak in a high voice at least; can't yet make it sound clearly feminine though.
Иди нaхyй cyкa блять.
Fucking kids and their top40.
Fuck off with that shit.
If I were to pick BGM for an office I'd choose classical over this crap.
>>742472684
I'm a shit singer though. As for speaking, my voice is a kinda soft unless I want to project it, then it's deep and loud, almost kingly. I can't really talk like a girl, but I can moan like one, heh,
>>742472844
Give parade commands.
>>742472944
Sounds gay. I'd rather act or voice act. I'm very good at faking emotion and getting into character.
>>742473018
Perhaps, if said emotion was "edgelord teenager".
>>742473056
Nope. Any kind of emotion. I'm a great actor, but I don't like to in front of people due to the stigma of Hollywood.
>>742473127
"I'm actually really good at doing something, except for in situations where it's expected of me to be good at it".
>>742473196
>huge ass
>facesitting
Damn. That's an old ass pic I saved, literally. I guess I had latent fetishes that I didn't realize until later.
>>742473244
I never said I wasn't able to, dumbass. I said I don't like to. I hate the entertainment industry.
i made a program that learned how to play metroid a while back prob one of my prouder moments me being in highschool at the time, I dont have the program so dont ask
>>742473672
Metroid would be neat. Did it already have a map already loaded in to it, and already knew all the locations of items; or did it have to explore itself and find it's own items? Also; NES Metroid?
>>742474086
gay
>>742474117
>117
You stupid, Chief?
>>742474156
no u
>>742474201
Nah you're pretty fucking stupid if you think that image is gay.
>>742474244
i was talking abot u fag
>>742474285
No you weren't. You're just a retard.
>>742474319
ur a retard
>>742474499
Not as retarded as you apparently. And I guess not as gay since you seem to obsess over it so much. Not like loving huge, women's asses is gay in the slightest.
>>742474600
degenerate
>>742474758
Says the necropedophile.
>>742474911
Morning, Ethereal.
>>742466019
high
https://www.youtube.com/watch?v=kc6kVjnUbCo
low
https://www.youtube.com/watch?v=uCUnhoe0rpU
and yes, i can throw my voice kinda like that
I listen to the bucket because i dont like lyrics generally
>>742475016
bleh
>>742475230
Same.
>>742475353
Morning, Winter.
>>742475400
Good morning.
fuckin, one thorn DESTROYS my last night.
and at least a half dozen micro nightmares during the second half of trying to sleep hours.
>>742475446
>>742475556
Never heard of it until now. Sounds pretty awful.
I want nice music to meditate to
Any music really, anything can be meditated to as long as it doesn't contain surprises.
Even Passenger of Shit music, because instead of unexpected screaming, it's continuous screaming.
Machine learning isn't real AI. To learn about it, read about it. Back-propagation (adjusting the weights of neuronal outputs based on the success/failure of the model) is how you train them. Machine learning is used to develop models (networks with assigned weights) that you then run. People don't generally train them on the fly because the training is more computationally expensive, and to do it well requires feedback on whether the results are right or not.
What, nothing to say about this one, Alex? >>742475341
>>742475657
Are you looking for some modernist palette or a more classical flavour?
https://www.youtube.com/watch?v=-XWmBeVRWoc
>>742475756
I dunno, I'm going to see if I'm right about Passenger of Shit being perfectly fine music for meditiation.
>>742475657
im thinkin this
https://www.youtube.com/watch?v=ZFkblAeoyMA
I just remembered another low one. Because of my vocal fry, my death growls are stupid low. Never heard anyone else able to do this.
https://www.youtube.com/watch?v=KuBAWEVz6Pw
>>742475890
heh, i cant go quite that low...
>>742475961
Maybe you could. Just takes practice. I've practiced manipulating my voice basically my whole life which is why I have a ridiculous vocal range.
>>742476046
they way i do my growl is weird, i actually inhale whilst vibrating my chords. i can get it pretty low, but its usually around Nathan Explosions tone. 'Cept he just does it naturally. and because i dont do it every day i cant do it for extended periods.
>>742476046
If you spent as much time practicing your saber techniques as much as you do your bullshit, you would rival Master Yoda as a swordsman.