>Our relationship to computers has changed, instead of programming them we now show them and they figure it out. That's a completely different way of using computers and computer science departments are built around the idea of programming computers and they don't understand that this "showing computers" is going to be as big as programming computers.
youtube.com
PROGRAMMERS BTFO
Other urls found in this thread:
arxiv.org
xkcd.com
vimeo.com
youtu.be
twitter.com
How is "showing computers" what to do not programming?
t. non-professional programmer
OP is retarded
>t. Newbie
>t. cucked programmer
It's the next level of programming. It's more like teaching.
"machine learning" == "shit's fast enough now that we can be lazy and inefficient"
>NNs won't need to be programmed
Ok
it's not nearly fast enough but it's getting there
Machine learning is basically pleb tier languages 2.0.
Pleb tier languages were for people who couldn't manage memory and other low level interfaces.
It allowed plebs to write computer programs, but they had to pay dearly with performance drops, random GC stops and a fuckton of increased memory usage.
This... this is worse. This is for people who can't even program. And the costs are even higher.
Also good luck with bug fixing. Instead of being able to start your debugger, find the line that's wrong and fix it, you need to change the sample data and let the computer spend days trying to recompute its model.
wow.
On a long enough time line all jobs become automated (provided we dont destroy ourselves) only thing going to be left is art and only as a novelty mostly.
I do post-doc research on operating systems at my country's top university. I make 207.000 DKK while your mommy brings you Hot Pockets and finds that you're still asleep at 5pm after crying yourself to sleep and sleeping for 16 hours.
How will this effect embedded programming, where, only recently has object oriented begun to be adopted.
t. Retards.
These articles are dog shit in conveying what the current status of 'teaching computers to teach themselves' actually is.
This stuff, in actuality, is over-glorified brute forcing. The software is absolutely incapable of abstract thought, which is a big deal. The software still needs to be told what the parameters for success is, as well, the tools the software utilizes must be developed as well. From there, it just uses those tools and aggressively brute forces its way to success. Once it succeeds, if it does, it will write the parameters that led to that success and then using some algorithm developed by a programmer, how to apply that to future situations.
Until the day comes (if it does) that computers are capable of abstract thought, this is all 'AI' will ever hope to be, just really over-hyped brute forcing. Someone still needs to be there to set the conditions for victory and give the software the needed tools.
Do you think capsules will work? I think they are the future of deep learning.
>This stuff, in actuality, is over-glorified brute forcing.
It's fucking retarded, nobody bothered to use computers for 10 years doing anything other than social network crap and all of a sudden someone was like "hey, remember that dumb shit we used to do in DOS with morphing faces that took days, let's pretend that there's all this neurally wurly shit going on and do stuff like that again, only the computers are faster now" I fucking hate the future
>only thing going to be left is art and only as a novelty mostly
Even then
vimeo.com
youtu.be
Bots can also write news articles too, how long until one writes a best selling novel I wonder.
Industrial machines already have a "teach" feature.
You still have to code that feature.