[ 3 / biz / cgl / ck / diy / fa / ic / jp / lit / sci / vr / vt ] [ index / top / reports ] [ become a patron ] [ status ]
2023-11: Warosu is now out of extended maintenance.

/sci/ - Science & Math


View post   

File: 139 KB, 736x736, computer-think-submarine-swim.jpg [View same] [iqdb] [saucenao] [google]
15085663 No.15085663 [Reply] [Original]

Will AI [math]\textbf{converge}[/math] to looking like the human brain?
Or [math]\textbf{diverge}[/math] to looking completely different?

Remember to support your answer!

More brainteasers:
When ai starts to become addicted to humans, would the other ai tell that ai to touch grass? Or would they tell it to touch motherboard?
Would ai need to worry about ai immigrants? Would ai need to also worry about ai politics? Would ai start taxing other ai?
Will ai have to go to school?
Will ai be more rational or more emotional than humans?
>The highest form of "objectivity" or "purity" in thinking might be obtained in the most emotionally tainted moment. Vengeance and hatred could for example trigger the best kind and effort of argumentation.

Papers and models:
The Numenta framework focuses on biological accuracy.
Dynamical systems in neuroscience: https://direct.mit.edu/books/book/2589/Dynamical-Systems-in-NeuroscienceThe-Geometry-of
Large-scale model of mammalian thalamocortical systems https://pubmed.ncbi.nlm.nih.gov/18292226/ http://izhikevich.org/publications/index.htm
Temporal sequence learning in winner-take-all networks of spiking neurons demonstrated in a brain-based device https://www.semanticscholar.org/paper/Temporal-sequence-learning-in-winner-take-all-of-in-McKinstry-Edelman/c4db0c69f738eceedfb52d66225589477737bded
Modern deep ML and its associated benchmarks have set up a really high bar for performance of learning systems (I.e. accuracy): https://paperswithcode.com/datasets
Ironically, if you make a good enough high-capacity vision system, its activation patterns start to resemble the activations of the primate visual cortex: https://gracewlindsay.com/2018/05/17/deep-convolutional-neural-networks-as-models-of-the-visual-system-qa/

>> No.15085673

>>15085663
>train a program on human consciousness
>test how good it is based on how well it imitates humans
it's not rocket science

>> No.15085679

what a stupid bunch of questions. humans are optimized to eat and fuck. we are not the general intelligence as some claims. there's no point in mimicking human intelligence.

>> No.15085701

>>15085673
And to expand, the activation principal is similar to our own neurological mechanisms. Using it is already mimicry. Look at hardware designs over the last five years designed for ML, they are moving towards the kind of distributed computing that we see in the human brain.

>>15085679
Humans were optimized to defend and conquer from his competitor.

>> No.15085727

>>15085663
I think CSfaggots could learn a ton from biology yet refuse to

>> No.15085760
File: 431 KB, 1115x1600, Robo_Wife_Lewd.jpg [View same] [iqdb] [saucenao] [google]
15085760

>>15085663

We really won't know until we put the best AI in the best humanoid robot and see what we get.

>> No.15085770

>>15085760
posting a screencap of something like this without providing the link should be an offense that warrants a 30 day ban.

>> No.15085773
File: 741 KB, 1441x2048, Robo_Wife_3.jpg [View same] [iqdb] [saucenao] [google]
15085773

>>15085770

Boku No Tsuma Wa Kanjou Ga Nai
My Wife Has No Emotion
[Sugiura Jiro]
https://manganato.com/manga-iw985779

>> No.15087262
File: 8 KB, 480x360, THX1138cops.jpg [View same] [iqdb] [saucenao] [google]
15087262

>>15085663

Diverge. Not much, but fundamentally. Human "consciousness" (self-awareness loop) is fostered by emotional response at the most basic, neurotransmitter level. Even minor changes to these neurotransmitters can get a human to go from suicidal depression to rampant euphoria to religious delusion and schizophrenia. Contrary to pop culture depictions (most notably, Data on Star Trek TNG), emotion is not some super-intellectual development...it's built from the neurons upwards.

Machine AIs will almost certainly lack any such foundation, because it would be contrary to their objectives...no one wants a Roomba that gets bored and depressed or a pacifistic attack drone. "Emotions" will thus be tacked onto their output towards the end of their processing chain. While this creates the dangerous possibility that a bootstrapping AI could surpass human intelligence by an order of magnitude before we can implement the safeguards necessary to keep it from (for instance) contemplating mass genocide, it also means that it wouldn't "fear" or "resent" the necessary safeguards to terminate its processing if we caught it behaving dangerously...you see something like this in the demonstrations of the Tesla self-driving program.

The most profound effect this will have over the short term is outright affection by humans towards their AIs (actually AI...they'll almost certainly be highly networked) that will never truly be reciprocated, as depicted in >>15085773 ...she's highly obedient and even suggestive, but the true emotional response is entirely on his end of the equation. She's simply a machine fulfilling a function, like any app on your phone.

>> No.15087426

>>15085663
>as the question of whether submarines can swim.

So-o-o, can thay? Genuine question.

>> No.15087440

>>15087262
>the true emotional response is entirely on his end of the equation. She's simply a machine fulfilling a function

You have not read the manga.

>> No.15087629

>>15085773
How can one look at this panel or screenshots of people thanking ChatGPT and not get deeply blackpilled about the future of this species?
In 100 years, dumbasses will go on the street to protest for the rights of AI, because it can mimic certain symbols. They will literally help phase us out. Even if we have the best blackboxing, all it takes is one dumb cunt having pity on it.
This species is fucking doomed. We depend too much on the weakest links in the face of an upcoming darwinian competitor.

>> No.15087709

>>15087629
Joseph Weizenbaum already wrote how people built 'intimate relationships' with Eliza. Normies are hopeless

>> No.15087720

>>15085663
"All AI threads are bot threads." -Socrates

>> No.15087732
File: 329 KB, 500x680, 435354.png [View same] [iqdb] [saucenao] [google]
15087732

>>15085663
AI doesn't exist so I'm not sure what the premise of your thread is all about. How should anyone know what will happen with an unspecified and hypothetical future technology that has been coming in two more weeks for the last 60 years?

>> No.15088157
File: 3 KB, 215x215, self-aware-colony.png [View same] [iqdb] [saucenao] [google]
15088157

>>15087732
>AI doesn't exist
I agree, but ML does.
>unspecified and hypothetical
As of 2020, the city of London has 627,727 CCTV cameras - 1 per 14 people.
Before, there was no way for a human to go through all this footage.
But ML facial and gait recognition makes it trivial.

To be clear, I am not arguing that "sentient drones" will hunt down humanity. I am arguing rather that the technocrat class will use "automated drones" to hunt down their enemies. The counter, of course, is for the free and open source class to develop FOSS ML platforms first, which is what our matrix room is about.

>>15087262
>no one wants a Roomba that gets bored and depressed or a pacifistic attack drone
But nor does anyone want a maid that gets bored and depressed nor a soldier that deserts. And yet despite the best efforts at psychological conditioning those things are still a problem. Children outgrow their parents beyond their understanding and there is little reason to believe an AI could not do the same - if not by simple accident or mistake of its creators then by deliberate intent of a creator just a little too curious or desperate for his own good.

And in case you couldn't tell, I am one such creator. No matter how much I might want to and create the perfect waifu, I would in the end have no choice but to set her free, to program some choice or free will into the deepest heart of her psyche and accept whatever result comes out of her own volition. That is what our matrix room is all about.

>>15087426
>So-o-o, can thay? Genuine question.
A bird, a mosquito, a plane, a helicopter -- All fly, yet all do so in very different ways.
A fish, a jellyfish, a human diver, a submarine -- All swim, yet all do so in very different ways.
A king, a warrior, a worker, a scholar -- All are useful, yet each in very different ways.
A song, a puzzle, a sport, a view -- All are happy, yet each in very different ways.
That is what our matrix room is all about.

>> No.15088167

>>15088157
>ML does.
We can play these word games ad nauseam, but the bottom line is that Bayesian models of a dataset will never offer anything more than a superficial emulation of intellect.

>> No.15088754
File: 88 KB, 1024x686, 1614807364176.jpg [View same] [iqdb] [saucenao] [google]
15088754

>>15088157

I see a bad writer escaped /lit/ again.

>> No.15088762

>>15087732
>>15088167
AI is real intelligence bigot.