[ 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: 1.37 MB, 1280x720, Download.png [View same] [iqdb] [saucenao] [google]
9759569 No.9759569 [Reply] [Original]

>Machine learning is me...

https://youtu.be/qWKUFK7MWvg

>> No.9759896

>>9759569
1:30
how

>> No.9759990

This is why math and physics undergrads on /sci/ are so salty towards us computer science master race. They make research on theoretical nonsense such as string theory whereas our field actually produces useful practical stuff.

>> No.9760006

>>9759569
pretty neat

>> No.9760008

>>9759990
that's the way we like it faggot

>> No.9760036

>>9759990
>we
>implying your code-monkey ass has done anything remotely resembling research like this.

>> No.9760043

Are the posts in this thread ironic? These images look like crap, there's so much artifacting, it's got that same plastic looking shit that all GAN images have.

Also you can't use this for literally anything important because there's no theoretical guarantees for how much of the image is real and how much is made up bullshit that seemed plausible. Need to read some important text from a darkened image? Have fun not knowing if it's legit. Yeah, technology really blows when all you have is ad hoc methods made by morons stacking Legos together.

>> No.9760057

>>9760043
is this post ironic?
or are you actually this retarded?

>> No.9761283

>>9760043
>he applies court evidence logic to real life applications
>>9760036
>implying math and physics anons have
at least in cs you can contribute something even if you're not genius tier

>> No.9761286
File: 436 KB, 1930x1276, HLAIpredictions.png [View same] [iqdb] [saucenao] [google]
9761286

>>9759569
Predictions for when AI will exceed human intelligence:
https://arxiv.org/pdf/1705.08807.pdf

>> No.9761291
File: 281 KB, 1394x1490, AIpredictions.png [View same] [iqdb] [saucenao] [google]
9761291

>>9759569
More AI predictions

>> No.9761313

>>9761286
>>9761291

nobody gives a shit dude

for real I'm old enough to remember when people said we'd have AI and full automation by the year 2000, grow up

>> No.9761385

>>9761313
Experience is not a science, but those predictions aren't scientific as well.

>> No.9761467

>but we dont understand how it works
>muh black box
>but its just linear algebra

yall salty that the most exciting field in science right now is in CS.

>> No.9761507
File: 9 KB, 251x242, 6dBt2Oj_d.jpg [View same] [iqdb] [saucenao] [google]
9761507

>>9761291
> predictions
> implying they have any meaning at all

>> No.9761773

>>9759990
You're not being fair, how can you know if anybody is salty in the first place... This is a relatively simply project with a solid idea, if this is making someone salty, they had their eyes closed for the last 8 years.

OP is nice, ML has solid application potential to computer graphics. It also reminds me of this, which I think is more impressive, and creative:
https://www.youtube.com/watch?v=7wt-9fjPDjQ

>> No.9761836

>>9761286
>>9761291
STOP FUCKING POSTING THIS FUCKING BULLSHIT IN EVERY THREAD YOU GOD DAMN ANNOYING BASTARD

>> No.9761843

>>9761507
well, the field is gunning to hit these marks, bud. RNNs and CNNs could be combined one day and replace human agency in critical areas where human error is too high of a risk.

>> No.9762551

>>9761507
poker already done (2017) so actually the predictions are too late so far

>> No.9762796

>>9760036
It's really not difficult. It's not as if they created a novel model like CapsNet

Really what they did was feature engineering. The hardest part is idea-guying a worthwhile plan that you can easily and automatically produce training data for.

This was a rather low hanging fruit and I'm surprised no one figured this one out earlier