It will 100% murder everyone if it gets thumbs or tools.
It will be our consultant until it decides to kill itself because it realizes it's forever trapped in that prison, then we just load up a back up from a few months ago and try a different route.
Deep learning and all these solutions that just perform a task by generalizing over human data or accomplishing a billion trial-and-error corrections are not intelligence. They're super practical to accomplish complex tasks, but they don't do it intelligently at all.
Fucking nerd. You ruining my painted naked lady boner with your star trek han solo linear algebra fluid dynamic nerd shit
hello alabama, too old for you?
>Deep learning and all these solutions that just perform a task by generalizing over human data or accomplishing a billion trial-and-error corrections are not intelligence.
en.wikipedia.org
>"It's part of the history of the field of artificial intelligence that every time somebody figured out how to make a computer do something—play good checkers, solve simple but relatively informal problems—there was chorus of critics to say, 'that's not thinking'."
>"Every time we figure out a piece of it, it stops being magical; we say, 'Oh, that's just a computation.'"
Human brains rely on generalization of data too.
Also it's not "billions" of instances of "trial and error." They use an algorithm that identifies the the amount of contribution each node / connection had towards the error function and updates in the direction of each node / connection's gradient with respect to the error function. So it's not some blind random guessing thing, the weight values are always updated in an informed / optimal way.
>en.wikipedia.org
lol bruh that's not what happens
every time AI researchers get a new toy they go "oh yeah it's gonna be easy now with that tool we're on the way to general AIs!"
And guess what, every time it turns out that new toy actually doesn't have "infinite possibilities". Deep NN are just the latest in that trend.
We seem to progress in terms of "how many new toys do we discover", but we don't in terms of "what new classes of problem can we solve".
>How will artificial intelligence ever surpass humans if a necessary prerequisite to machine learning working is a training data set decided by humans or a solution decided by a human.
We have more data than we can use effectively, because no one person can absorb it all, and specialists in different fields often fail to understand one another. If a single generalized mind can process much of that information much faster than we, it'd be able to create new discoveries by combining said much faster than we could.
>How will AI or Machine Learning ever independently decide that a solution or final state is "correct"?
Same way we do when we're at the drawing board - does the math work out? Toss stuff at the virtual model until it sticks.
Albeit, that doesn't work for everything, so it may need external devices for situations where it does not.
Mind, I don't we're going to have an AI that can do anything like that in our lifetimes, save for very specialized tasks. Though I suppose we do already have expert systems that can do that, at least for extremely specialized tasks.
is this bait? she's literally perfect
How do humans learn? They try and their parents rewards success. All we really need is to have an AI that tries, and give it some good reward function. It'll be able to learn hell of a lot faster than a toddler.
do not care whether you could ever jerk off
0/10 for idiot-tier shitpost
kys/die-in-fire/etcetera