[ 3 / biz / cgl / ck / diy / fa / ic / jp / lit / sci / vr / vt ] [ index / top / reports ] [ become a patron ] [ status ]
2023-11: Warosu is now out of extended maintenance.

/sci/ - Science & Math

View post   
File: 77 KB, 670x446, robot-team670a.jpg [View same] [iqdb] [saucenao] [google]
7280623 No.7280623 [Reply] [Original]

>Muh only narrow AI
>Muh no progress towards general AI
>Muh we need magic new technology

No you faggots, it's deep learning all the way. But you'll keep making those retarded exucses until the killdozers push your luddite home into the ocean.

http://newscenter.berkeley.edu/wp-content/uploads/2015/05/robot-team670a.jpg

>> No.7280625

>>7280623
http://newscenter.berkeley.edu/2015/05/21/deep-learning-robot-masters-skills-via-trial-and-error/
and here's the link to the article.

>> No.7280634

>>7280623
nude of the girl ?

>> No.7280648

>>7280634
this

>> No.7280656

>>7280623
deep learning is powerful and promising but it's absolutely not "strong ai"

deep neural nets are also really finicky and hard to train, and they tend to pick up strange behaviors

>> No.7280658

>>7280623
Stop jumping on the hype bandwagon called deeplearning.
If you have ever done training any deep net, you would have realized that it's stupid as fuck. It was fucking dead in the 1950s 1960s then revived recently only because of better hardwares.
It will become stagnant soon and you will realize how stupid you are.
>Multiple matrices multiplication
>General AI
lol

>> No.7280672
File: 3.56 MB, 300x200, ourRoboticOverlords.gif [View same] [iqdb] [saucenao] [google]
7280672

>>7280656
>they tend to pick up strange behaviors
so just like us then

>> No.7280673

>>7280623
T-this looks like a porno... Please tell me it is.

>> No.7280676

>>7280656
>>7280658
So how do we come up with a better solution?

>> No.7280679
File: 1.60 MB, 500x375, CnSTbEk.gif [View same] [iqdb] [saucenao] [google]
7280679

>>7280672
robot shaming thread?

>> No.7280689

>>7280623
the only progress currently being made towards strong AI is the work done at Deepmind

others can suck muh dick

>> No.7280690

>>7280676
The answer is obvious: we don't know yet.
If it was realized then you would have seen general AI every fuckingwhere.

>> No.7280700

>>7280672
Thanks. That gif gave me a good laugh.

>> No.7280703
File: 984 KB, 416x234, 1402835513414.gif [View same] [iqdb] [saucenao] [google]
7280703

>>7280679

>> No.7280717

>>7280672

Awww, they're like children having tantrums...

>> No.7280727

>>7280703
>>7280672
>>7280679

kek'd

>> No.7280729

Why did they make him look like a cash register? Fuck if I were him I WOULD kill them all for doing that shit to me.

>> No.7280732

What is a woman doing there? They try to give her intelligence?

>> No.7280733
File: 2.91 MB, 608x421, bipedal motion.gif [View same] [iqdb] [saucenao] [google]
7280733

>>7280703

>> No.7280739
File: 163 KB, 2048x1536, bb.jpg [View same] [iqdb] [saucenao] [google]
7280739

Deep...

>> No.7280756
File: 1.94 MB, 250x184, 7VYD6E5.gif [View same] [iqdb] [saucenao] [google]
7280756

>>7280733

>> No.7280782

>>7280756
lewd...

>> No.7280784

>>7280732
kek

>> No.7281276

I'm pretty sure they just did the same thing Deep Mind's Q-Network but with a real robot. Which is actually impressive, but Q-Networks stop scaling very quickly, as evident in Deep Mind's Atari experiments.

>> No.7281309

>>7280679
Laugh it up while you still can, humans.

>> No.7282420

>>7281276
Deep net is hitting a wall. The current abstraction ability is still pretty primitive (i.e. using a huge net with the best optimization techniques, it iterates for numerous trial and errors in real experiments, hence it would take a really long time for a real robot to repeat some actions until it reach an optimal value). We can make it faster by using better params but that's it. The current techniques are not scalable to "general AI" or "Strong AI".

>> No.7283865

>>7280672
KEKEKEK