/sci/ - Science & Math » Thread #12455781

138KiB, 635x355, cognition-article.jpg

View Same Google iqdb SauceNAO

Anonymous Sat 12 Dec 05:18:45 2020 No.12455781 View Reply Original Report

Quoted By: >>12455801 >>12456033 >>12456036 >>12456214 >>12456417 >>12458723

I keep seeing 'AI' threads posted here, but the fact of the matter is we don't have anything I'd call 'AI'. When I say 'AI' I mean *general* AI, not the garbage they have right now.

But we can't have General AI until we define what THINKING/REASONING/COGNITION actually *are*.
Can you define those qualities?

Anonymous

Anonymous Sat 12 Dec 2020 05:25:43 No.12455801 Report

Quoted By: >>12456019 >>12456238

>>12455781
>When I say 'AI' I mean *general* AI, not the garbage they have right now.

So things are only 'really smart' if they think like human do?

Anonymous

Anonymous Sat 12 Dec 2020 05:29:37 No.12455815 Report

Quoted By: >>12457781

Yes pop science and the media have done a poor job at characterizing AI. The current iteration of AI is used in marketing and it’s really just a deep learning/machine learning crystal ball used to analyze and predict purchasing habits. A republican male who owns a F-150 pickup is probably likely to purchase a holster. Even Africans can make this connection. But why are single father bald male condominium owners incredibly likely to own blue sedans? No one knows. But the algorithm sees the connection and it’s a Tangible quantification of human behavior. Do we really have free will?

Anonymous

Anonymous Sat 12 Dec 2020 05:35:22 No.12455829 Report

Quoted By: >>12456027

Thinking is the ability for the brain to produce novel content by synthesizing learned material. Reasoning is simply finding a compromise between two parties, regardless of whether or not one of the parties is imagined. Cognition is a bit trickier. It's an amalgamation of awareness and knowledge and the ability to fluidly interact in respects to these two concepts.

Anonymous

Anonymous Sat 12 Dec 2020 07:21:03 No.12456019 Report

Quoted By: >>12456037

>>12455801
>So things are only 'really smart' if they think like human do?
Yes.
What they have right now is crap so far as I'm concerned.

Anonymous

Anonymous Sat 12 Dec 2020 07:22:15 No.12456027 Report

Quoted By:

>>12455829
Those are broad generalizations, and while accurate, they don't explain *how* these things work. That's what needs to be figured out.

Anonymous

Anonymous Sat 12 Dec 2020 07:23:58 No.12456033 Report

Quoted By: >>12457550

>>12455781
>the ai right now isn't like in the scifi movies i jerk off to so it's garbage
who let reddit on this bort?

Anonymous

Anonymous Sat 12 Dec 2020 07:25:10 No.12456036 Report

Quoted By: >>12457681 >>12457781

>>12455781
Is anyone else incredibly disappointed by AI in video games? It's almost like they haven't made any advancements for decades

Anonymous

Anonymous Sat 12 Dec 2020 07:25:24 No.12456037 Report

Quoted By: >>12456040 >>12457554

>>12456019
What are your thoughts on GTP-3 then?

Anonymous

Anonymous Sat 12 Dec 2020 07:28:08 No.12456040 Report

Quoted By: >>12456059

>>12456037
Meme chatbot.

Anonymous

Anonymous Sat 12 Dec 2020 07:38:17 No.12456059 Report

Quoted By: >>12456081 >>12457293 >>12457562

>>12456040
See, that's what I thought. You have no appreciation for the level of advancement that has actually been achieved and have absurdly high standards for what needs to happen in order to care at all.

You can't define 'thinking' so you assume that what your own brain does is much less 'automatic' than it is because it happens to be connected to your own consciousness, which provides feelings of understanding and willfulness to you.

Anonymous

Anonymous Sat 12 Dec 2020 07:48:06 No.12456081 Report

Quoted By: >>12456256

>>12456059
>advancement
Throwing large data into a powerful computer and getting good model isn't called "advancement".

Anonymous

Anonymous Sat 12 Dec 2020 09:14:41 No.12456214 Report

Quoted By: >>12456362

>>12455781
>When I say 'AI' I mean *general* AI,
Ain't that adorable.
But who gives a fuck what you mean when you say it? Other people use the term to refer to machines that think. Get the fuck over yourself and fuck off with your worthless philosophy.

Anonymous

Anonymous Sat 12 Dec 2020 09:24:30 No.12456238 Report

Quoted By: >>12457571

>>12455801
Humans are not general intelligences. We are programmed for a few extremely specific tasks and our behaviour is a complex side effect of those.

Anonymous

Anonymous Sat 12 Dec 2020 09:33:16 No.12456256 Report

Quoted By: >>12457781

>>12456081
Then what is? If something demonstrates fast, few shot learning (implying the ability to heavily bias itself), is that enough? Does it have to be able to handle a large enough variety of subjects? Does it have to be conscious? What if some computer could actually be shows to be conscious, but was a complete dumbass on par with a retarded child? Would that be 'General AI'?

Anonymous

Anonymous Sat 12 Dec 2020 10:34:11 No.12456362 Report

Quoted By: >>12457781

>>12456214
But they aren't thinking.

Anonymous

Anonymous Sat 12 Dec 2020 10:54:55 No.12456417 Report

Quoted By: >>12457571

>>12455781
>not the garbage they have right now.
You mean statistics? Are you saying statistics is garbage? You wann fite me. bitch?

Anonymous

Anonymous Sat 12 Dec 2020 17:49:35 No.12457293 Report

Quoted By: >>12457781

>>12456059
GPT-3 is the biggest meme though. Like dude if your machine learning research involves having billions of parameters then you are just brute forcing. Like dude its like saying that having 60 GPUs in parallel working on brute forcing a password is a breakthrough in cyber attacks. Few-shot learning is still absolutely garbage

Anonymous

Anonymous Sat 12 Dec 2020 19:13:08 No.12457550 Report

Quoted By: >>12457781

>>12456033
I want GENERAL AI, not the meme 'deep learning algorithms' they have now. None of those have any capacity to THINK, no personality. Can't get those things until you can define how they work in biological brains.

Anonymous

Anonymous Sat 12 Dec 2020 19:14:12 No.12457554 Report

Quoted By:

>>12456037
>What are your thoughts on GTP-3 then?
It's just an Eliza program on steroids.
There was a demo of it I tried. It was nonsensical. There was no way you'd mistake it for a human.

Anonymous

Anonymous Sat 12 Dec 2020 19:17:22 No.12457562 Report

Quoted By:

>>12456059
The other two who already responded to your rant more or less covered things.
Meanwhile you're avoiding the subject because you have nothing of value to add to the conversation of the original topic.
Also you sound like a 'fake AI' fanboy like too many. I wouldn't at all be surprised if you secretly think that if they throw enough hardware and """training data"" at them, they'll suddenly 'wake up' like Mycroft in 'The Moon Is A Harsh Mistress'. Ain't happening.

Anonymous

Anonymous Sat 12 Dec 2020 19:18:46 No.12457571 Report

Quoted By:

>>12456238
Uh, No, that's incorrect.
>>12456417
I believe the quote is 'Lies, damned lies, and statisitics'. Also 'Internet Tough Guys' don't impress me.

Anonymous

Anonymous Sat 12 Dec 2020 19:22:28 No.12457578 Report

Quoted By: >>12458402

> While it is conceivable in theory that a machine could be more intelligent than a human being, today’s technology only covers a tiny area of our thought, namely the recognition of patterns. Also, there is hardly a single trend in biotechnology that is progressing exponentially. AI is also an energy-guzzling process. The brain works with 20 watts. This is enough to cover our entire thinking ability. AI needs an incredible amount of energy to recognise a picture of a penguin from 10 million images. To solve the problem, it requires entire data centres that need to be kept cool. If we wanted to use AI to reproduce everything the human brain is capable of, we would need a vast number of nuclear power plants to provide the necessary energy, assuming it would even be possible that is.

Anonymous

Anonymous Sat 12 Dec 2020 19:39:08 No.12457637 Report

Quoted By:

A GAI should be able to solve this
https://www.kaggle.com/c/abstraction-and-reasoning-challenge

Anonymous

Anonymous Sat 12 Dec 2020 19:53:38 No.12457681 Report

Quoted By:

>>12456036
What are some of the more "advanced" ones you've seen? Alien: isolation is the only one I know of that looked interesting.

Anonymous

Anonymous Sat 12 Dec 2020 20:23:40 No.12457781 Report

Quoted By: >>12458392

>>12456256
Don't bother hes a dogmatic retard

>>12457293
The brain has more parameters than GPT-3 has, shut the fuck up retard.

>>12455815
That clearly isnt "all" its used for though. You have reinforcement learning models and language models that can be used on a huge variety of tasks.

>>12457550
Shut the fuck up retard. If we had those things we'd LITERALLY HAVE SOLVED THE ENTIRE FIELD. We're getting there gradually. You're just being a child.

>>12456036
Because using neural nets in games with a large number of agents would be a processing hog.

>>12456362
Define think, retard

Anonymous

Anonymous Sun 13 Dec 2020 00:30:18 No.12458392 Report

Quoted By: >>12459805

>>12457781
>Shut the fuck up retard. If we had those things we'd LITERALLY HAVE SOLVED THE ENTIRE FIELD. We're getting there gradually. You're just being a child.
What? Fuck you, that's why I made this entire thread you asshole, to DISCUSS the subject!
You belong in >>>/b/ not /sci/, all you're doing is insulting people and being a dick. Fuck off, you're not wanted here.

Anonymous

Anonymous Sun 13 Dec 2020 00:33:36 No.12458402 Report

Quoted By: >>12458720 >>12458727

>>12457578
>If we wanted to use AI to reproduce everything the human brain is capable of, we would need a vast number of nuclear power plants to provide the necessary energy, assuming it would even be possible that is.
Exactly this. They keep throwing more and more hardware and more and more data at it, seemingly in the hopes that it'll 'wake up' and suddenly start thinking. The approach is wrong, or at the very least woefully insufficient.
I believe that the answer lies in neurological research, aided by physics and engineering, and not with computer science alone.

Anonymous

Anonymous Sun 13 Dec 2020 02:31:38 No.12458720 Report

Quoted By:

>>12458402
This, patching computer chips together and forcing an AI to read Wikipedia obviously isn’t the way. We need more vomiting density first.

Anonymous

Anonymous Sun 13 Dec 2020 02:32:14 No.12458723 Report

Quoted By: >>12459400

>>12455781
AI is just rebranded fuzzy logic.

Anonymous

Anonymous Sun 13 Dec 2020 02:32:55 No.12458727 Report

Quoted By: >>12459400

>>12458402
This, patching computer chips together and forcing an AI to read Wikipedia obviously isn’t the way. We need more computing density first.

Anonymous

Anonymous Sun 13 Dec 2020 07:14:34 No.12459400 Report

Quoted By:

>>12458723
I don't disagree with you at all on that.
>>12458727
What do you mean by 'computing density', since on the surface it sounds like you're saying the opposite of what I already said?

Anonymous

Anonymous Sun 13 Dec 2020 07:17:32 No.12459407 Report

Quoted By:

UR ALL FUCKING RETARDS LMAO
t. harvard phd student in ML

Anonymous

Anonymous Sun 13 Dec 2020 10:48:37 No.12459805 Report

Quoted By:

>>12458392
Shut up. You clearly don't know anything about modern AI, and you clearly don't want to learn. All you wanna do is meme about how the field is useless because it literally hasnt solved every problem in existence.

Capcode	All Only User Posts Only Moderator Posts Only Admin Posts Only Developer Posts
Show Posts	All Only With Images Only Without Images
Deleted Posts	All Only Deleted Posts Only Non-Deleted Posts
Ghost Posts	All Only Ghost Posts Only Non-Ghost Posts
Post Type	All Only Sticky Threads Only Opening Posts Only Reply Posts
Results	All Grouped By Threads
Order	Latest Posts First Oldest Posts First

Your latest searches