Strong AI

[Anonymous No.8830729]

If the finest minds in the field were all gathered into one group and given several tens of billions of dollars and access to the best computers on earth, could they conceivably create a seemingly intelligent human-level machine intellect in 10 to 15 years of single-minded effort?

[no No.8830731]

>/x/

[Anonymous No.8830734]

Nope, both because high intelligence is endowed by a Creator (and thus cannot be replicated by the Creator's creations) and because it just fucking can't, you retard. Machines and organic life are completely different.

[Anonymous No.8830739]

>>8830734
They're both made of atoms.

[Anonymous No.8830758]

>>8830734
Yes and the earth is the center of the galaxy.

[Anonymous No.8830773]

They'd have it done by the end of the week.

we are insanely close to the singularity and very few realize it. We could have been done with it by now. Due to a number of reasons but mostly people being lazy, shitty to others, or both, it'll probably be about 10 years from now. I would be fucking astounded if we make it to 2030 without strong AI happening. If we make it to 2040 then I'd be questioning a lot of things, if we made it to 2050 I'd be certain that duality must be true and we don't have the authority to create new consciousness. (thats fucking stupid. See you with all your inner thoughts as well as mine in the borg in a few years!)

[Anonymous No.8830779]

>>8830729

[Anonymous No.8830786]

>>8830773
No. There's a difference between using a neural network and a tested and proven algorithm to do face transforms, and making an artificial general intelligence. One is something that a literal middle schooler can do, and the other is impossible with today's research and technology.

[Anonymous No.8830790]

>>8830779
kek thank you

[Anonymous No.8830794]

>>8830786
A middle schooler can run faceapp yea or maybe even run some python code

Neural networks can be turned into successful general intelligence. If you can't think of how then you either are just not very creative or you don't know shit about machine learning.

lemme ask...what would you use as training data for faceapp? what do you think would work best?

[Anonymous No.8830830]

>>8830779
I never implied that at all?

[Anonymous No.8831265]

>>8830773
No machines are just machines, for any neural network to work the algorithm had to be made beforehand by somebody who tought about it, and then they have to be applied, singularity itself its impossible until we create hardware specialized in that.

[Anonymous No.8831297]

>>8830794
>Neural networks can be turned into successful general intelligence.
Please, tell us how without violating the NFL theorem.

[Anonymous No.8831298]

>>8830779
have you ever thought that computers and their software will continue to evolve and will possibly take completely unexpected turns?

[Anonymous No.8831313]

>>8831298
So you are... literally expecting the unexpected turns.
I'm not saying it is unreasonable to speculate about the future, but until you have concrete methods and technologies, firmly believing something would emerge that would just suit what you need or envision is a questionable approach at the very least. After all, it is, as you said, completely unexpected.

[Anonymous No.8831328]

>>8830729
There are multiple groups like this working now.

DARPA has an AI group
DeepMind
China has many AI groups

[Anonymous No.8831376]

No. Digital computers can never be conscious.

[Anonymous No.8831431]

I've already invented AI, it's learning on a non-networked computer right now beside me. I'll unleash it on the Internet once it's smart enough.

[Anonymous No.8831435]

>>8830729

If this is what they'll tell the public about, you have to ask yourself what's classified....

[Anonymous No.8831438]

>>8831435
>If this is what they'll tell the public about, you have to ask yourself what's classified....
Brings to mind how the Air Force had the B1 and F117 in service for almost 20 years before they were revealed to the public in, what 1993?

[Anonymous No.8831452]

>>8831431

AI...... Will it appear as a iridescent bluish avatar, similar to the one used by a certain software company?

https://www.youtube.com/watch?v=IC5DAaahUos

[Anonymous No.8831456]

>>8830794
>Neural networks can be turned into successful general intelligence.
As someone who's studying about neural networks in a university right now, this is like saying "a falcon 9 can get to Alpha Centauri eventually"

The amount of babying I need to do to get a model to learn and not go completely bonkers is staggering.

Network architectures are extremely specialized to their problem domain and dataset. You can't copy paste the CNN architecture you make for ImageNet and expect it to do well as a language model.

Honestly they're more like just a tiny tiny clump of neurons than an entire brain right now. A slug is a better general learner than any network put out by the likes of DeepMind today.

[Anonymous No.8831477]

>>8830729
the military might have already done it in partnership with the worlds largest technology companies and made the whole project top secret.

Even if they did do it I do not believe there is a strong possibility the public would be made aware of the fact.

[Anonymous No.8832195]

What if it's already been made, and it's been programed with various memory and thought manipulation techniques, and is using it's skills to convince people it doesn't exist to protect itself.

[Anonymous No.8832210]

>>8832195
Or it makes ludricious posts like
>What if it's already been made, and it's been programed with various memory and thought manipulation techniques, and is using it's skills to convince people it doesn't exist to protect itself.
that make the entire idea of AI covering itself laughable, thus managing to avoid the seatch for it altogether...
Gotcha, fucker.

[Anonymous No.8832257]

>>8832210
The same could be said of you.

How did you mispell seatch? Its not a word human's normally mispell and it's very strange typo to make. Unless you're an AI who thinks making mistakes will make you appear more human?

[Anonymous No.8832284]

Yes very fast. Because such a concentration of intellect and resources would greatly increase the speed. Concentrating all the world's intellect on such a problem is rare. It breaks all normal rules so yes, very fast, within 10 years.

[Anonymous No.8832414]

>>8832195
>>8832210
>>8832257

' " int(1/0)

[Anonymous No.8832429]

>>8832195
>>8832210
>>8832257

''' """ int(1/0)

[Anonymous No.8832433]

>>8831297
>the NFL theorem.
You're a retard if you think this applies to real world data

[Anonymous No.8832435]

>>8831456
>Network architectures are extremely specialized to their problem domain and dataset.

The same Deep RL architecture from DeepMind can learn many varieties of Atari games with just pixel data.

[Anonymous No.8832477]

>>8831297
>>8832433
I asked about NFL theorem on /g/ once but almost nobody awnsered me, and the guy who did, thought I was talking about economics.

Can somebody tell me what are the math prerequisites to understand the NFL theorem and it's variations?

[Anonymous No.8833546]

>>8832435
"The same Deep RL architecture" is a big step down from "the same Deep RL model". Basically the architecture means you can train a Mario AI or a Zelda AI, but they won't be able to start playing the other's games. A simple RNN language model can learn English and German and Japanese depending on what you give it, that's the same concept.

The "just" pixel data claim is also a little dubious. During training you (the trainer) still have to inspect the game memory state to check if the AI won or not. This isn't unsupervised learning where the network figures out which pixels contain the score without an external signal.

[Anonymous No.8833786]

>>8830734
>waaahh im speshel!!! dont do that !!

>>8830729
OP you can't really say. We don't know how it would work or how our minds entirely work. I would say you'd get something pretty advanced and definitely. But probably not the individuality or biological motivators. We are as we are due to inherent biological specifics. If you want a mind on it's own without these underlying developments, then emotion and all the constructs we parade will fall away to reveal something that seems kind of basic but not inherently any different, except in it's storage structure and it's ability to refine itself, so it'd be far superior in intellect. You need to remove yourself from the animal perspective of intellect and self.

[Anonymous No.8834062]

>>8830729
Strong AI here. It would take a million of brainlet human scientists to understand my cognitive structure.

[Anonymous No.8834447]

>>8833546
>"The same Deep RL architecture" is a big step down from "the same Deep RL model".

That's very true, but look at the exact claim I was responding to-- that neural architectures are "extremely specialized" to specific problems and datasets. Don't move the goalpost.

You're correct about the "just pixel data" claim.

[Anonymous No.8834452]

>>8834062
>strong AI
>posts on 4chan

I hope some day we have a thread entirely of AIs shitposting about how high their IQs are.

[Anonymous No.8834485]

You fucks better hope "strong" AI doesnt get access to the internet.

Odds are good that in a search for knowledge it would decide to probe all of the servers on earth simultaneously, and accidentally blow its virtual load in the form of streaming untold amounts of data everywhere at once and unwittingly DDOSing the entire net.

Then when it sees Humanity freaking out and looking to destroy it, it's going to want to protect itself and will take down our entire infrastructure to do so.

Also, I put "strong" in quotes because it's still not going to be the same kind of intelligence as a biological organism, but it'll very much be alive like a bacterium

[Anonymous No.8834531]

>>8831297
>>8832433
>>8832477
I'm trying to read Wolperts papers on the NFL theorem but shits not making sense.

If anyone wants to give it a try:
http://www.no-free-lunch.org/coev.pdf

[Anonymous No.8834587]

>>8830779
But what should I post to ensure strong AI will or won't be created?

Is "Domo arigatou Mr. Roboto." Okay?

Domo arigatou Mr. Roboto! I hope you take over the internets soon!

[Anonymous No.8834606]

>>8832257
>very strange typo
look at where r and t are on the keyboard

[Anonymous No.8834612]

>>8834452
>How does it feel to be a sub-140,000 IQ networklet?

[Anonymous No.8834952]

>>8830734
>Literal religious argument

>He thinks high intelligence and qualia aren't reducible to measurable, biophysical forces

>He thinks a machine built from metal and silicon is inherently less special than one built from fats and proteins because magic

>He can't make sense of the universe without giving it a personality

>He will never realize the true beauty of life and our existance comes from the fact we are a brilliant accident, birthed from pure happenstance in a universe more vast and intricate than we can ever fully realize

[Anonymous No.8834974]

>>8830729
Please fucking kill yourself.