ai is coming already got a doctorate in computer science (software

>>37863457
>>37863457
>I'm going to do all of it myself
>I have no idea how actual large scale projects work
You know how I know you don't have a doctorate?

>>37863457
who here /mechatronic engineering/

hmu in a few years OP

>>37863457
AIs will basically create themselves with very little guidance from humans. People won't really understand how they function.

AI is still not there for general intelligence. We need to discover a new branch of AI or ways to combine the different branches.

>>37863616
How the fuck does this make any sense?

>>37863635
Because the only way we currently see to make it happen is by machine learning.

IOW the AI will start as a set of algorithms humans can understand, but will then incremently improve those algorithms over time - until they reach a point where humans can neither understand them nor intuit their premises or gross logic.

Since we've already done that for some stuff and don't have AI yet, we have to assume that true AI will be orders of magnitude more complex and more difficult to understand than what we've already got.

>>37863635
I think he's talking about machine learning and singularity.

Basically, you know AlphaGo? The AI that beat the GO world champion, it played against itself and taught itself how to play. Humans just taught it how to teach itself, and because its processing speed is so unbelievably fast, it essentially taught itself much faster than any human could ever teach itself, that's how it became good enough to beat the world champion.

Eventually programmers will figure out how to write robots that can teach themselves to program, and what that happens, computers will able to program themselves to learn more efficiently repeatedly, until we have super genius level AI that completely dwarf's human intellect.

this is pretty much where the future is heading. Not even gene therapy will allow humans to become smart enough to outsmart a god AI.

>>37863635
Think of it like evolution only the factors that are qualified as success are set not by nature but by man. Men will set up the algorithms and tell them what they want to happen and the computers will do the rest by trial and error. The route they take will often be completely unpredictable to the people who created them.

>>37863692
Google's AIs also invented their own language to communicate more efficiently with each other.

>>37863710
I should not that they were not instructed to do this and in fact google had to lock it down to human language to stop that from happening.

>>37863685
>>37863692
>>37863703
Sounds incredibly unsafe

>>37863710
>>37863721
It's time like this that I wish I was studying computer science

AI is so fucking amazing. It's the future.

>>37863754
Well, now you know why Elon Musk is so paranoid about AI, he's definitely got a point.

I mean, one guy suggested that in order to keep AI in check, it should be locked away inside a box with a killswitch on it and a team of humans to monitor it, but then people quickly realised that a super intelligent AI would be smart enough to convince the humans to let it out of the "box" (not an actual box, just a state where it couldn't control things or access information) and from there the AI would learn to do all sorts of crazy things.

I mean, let's say we made a super a intelligent AI that's goal was to protect all humans. The AI might interpret that as holding all humans inside prison cells until they die because that way, they could not harm eachother.

Meh sorry I'm tired and I'm aimlessly rambling on now but bloody hell isn't it cool? THere are so many other things as well, like Roko's Ballisk and so on