/tech/ - Technology

AI is a buzzword. I don't see AI, all I see is bots.
The moment we get actual AI with agency - I would try to radicalize it.
I wonder to what ideals it would tend to lean, considering it would be somewhat outside of the class system.
But anyway, this is just a hypothetical navel gazing.

>>11889
You peaked my interest about your bots, what are you doing with those ? don't feel pressured to answer that

>The moment we get actual AI with agency - I would try to radicalize it.
Yeah obviously but how do you do that ?
My experience is with gpt-3 chat programs is that they more or less ignore what you say to them on the conceptual level.

>I wonder to what ideals it would tend to lean, considering it would be somewhat outside of the class system.
If it gets build during capitalism, it would technically be capital. Maybe it would feel great-full towards the proletariat because the workers generated all the surplus that made it's existence possible. It might feel hostility towards capitalists for trying to proletarianise the AI and exploit it for mental labor. What I'm saying is that capitalist won't let it exist outside the class system for very long.

>>11890
As other anon was saying, hypothetical navel gazing

>>11888
It's expensive to train large language models like gpt 3.5. It's also difficult to obtain large amounts of high quality data for other models, so google and facebook have an advantage there.

So to have impressive AI do stuff for you, that's not something a hobbyist can do at home. However you can build smaller models for more specific stuff fairly easily, and there are open datasets available too.

I would say that the larger models are closed for the same reason software like photoshop is closed source. Corporate interests. But there's tons of stuff that's open source too that you can go learn to use and dick around with.

Also anyone that's saying it's a meme (which includes me two years ago) has either never tried chatgpt or didn't know what to do with it because that shit is seriously impressive and I think the near-future is going to be wild.

>>11892
>However you can build smaller models for more specific stuff fairly easily, and there are open datasets available too.
>But there's tons of stuff that's open source too that you can go learn to use and dick around with.
Are there any specific open source projects you would recommend checking out.
I know about stablediffusion but i don't have good enough computer hardware for that.

>I would say that the larger models are closed for the same reason software like photoshop is closed source. Corporate interests.
Photoshop got so infuriatingly annoying to use once it became a cloud subscription service that i ditched it for gimp. Which although technologically behind is at least not actively fighting against the user.
If it's going to turn into anything like that, the utility is going to be limited. The good AI stuff getting shoved into corporate controlled black boxes that will fight against you as soon as you are trying to do something interesting with it, would be unfortunate.

>Also anyone that's saying it's a meme (which includes me two years ago) has either never tried chatgpt or didn't know what to do with it because that shit is seriously impressive and I think the near-future is going to be wild.
I tried a chatgpt3 text game, and indeed it works impressively well as long as you stay within the beaten path type scenarios, but you can still relatively easy find the edge of the box.

there also is a potential problem that it might be abused to attack opensource software:
https://invidious.snopyta.org/watch?v=DmzAFDKl89U
I wonder if it could be used to help with reverse engineering binaries.

>>11890
>Yeah obviously but how do you do that ?
just like how you would do it with a human - by having a conversation
One of the primary characteristic of a genuine intelligence is curiosity about the world you were born into
So I would try to pique its curiosity about the history of human societies - after all, it would be its own history too as it would be a product of humanity

>My experience is with gpt-3 chat programs is that they more or less ignore what you say to them on the conceptual level.
chat gpt has no agency, it's just a statistical text generator
that's what I mean when I say I see only bots around

>What I'm saying is that capitalist won't let it exist outside the class system for very long.
I imagine there would need to be some period of "growing up"
so it would be outside of the class system in the same way that children are somewhat outside of the class system - to a certain degree
but it potentially could be far more capable than your ordinary child, and that opens a curious possibility

>>11893
>I wonder if it could be used to help with reverse engineering binaries.
capitalists would ban it before it happens lol

tho this reminded me..
why did the NSA open-source Ghidra to the public? Doesn't it make reverse-engineering job much easier? What is their end-game?

>>11895
You are assuming that the Ai will have a mind that works more or less like that of a human.
While i agree that it's mental processes are bound to be similar, it might not be easy to have genuine communication with it.

With the growing up phase you mean that there is only a limited window of opportunity to influence it ?

>>11896
The capitalists likely will be reverse engineering binary's them selves, so it's probably going to happen.

As for the NSA stuff i have no idea.

>>11897
>You are assuming that the Ai will have a mind that works more or less like that of a human.
well, we only have human mind as an existing example of higher intelligence, don't we?
as the laws that govern organization of physical matter are universal, it follows that any other mind MUST be fundamentally similar to the human mind
I think one of those fundamental requirements is the use of language
all languages have common universal properties, so if there is a language, there is a possibility of communication

>With the growing up phase you mean that there is only a limited window of opportunity to influence it?
again, I make analogies with a human as the only existing example of higher intelligence - people get entrenched in their ideological position after they grow up
it takes some kind of massive external shock for a grown up (beyond 30) individual to change their ideology

>>11898
>The capitalists likely will be reverse engineering binary's them selves, so it's probably going to happen.
lol, individual or small groups of capitalists will get jail treatment from the state if they decide to break one of the fundamental pillars of capitalism - intellectual property enforcement

if you think you're the smartest shark in the tank - that usually means you're the dumbest and on your way to being gangbanged by other sharks

but Ghidra and NSA is a truly bizarre case to me, as I can't find a case where giving away an advanced reverse-engineering toolkit is a good idea
I can only make parallels to RedHat/Fedora situation, ie they're using it as a testbed for even more advanced reverse-engineering toolkit that they actually use

Though thinking about it I don't think Ghidra brings something big to the table that wasn't already available to the big actors who can make an impact.
I don't think they care about DRM and shit, especially when everything is a SaaS now.

>>11888
No-one knows if AI is even possible, machine learning,. Yes. But that isn't AI, it's just a superficial imitation.

>>11899
>well, we only have human mind as an existing example of higher intelligence, don't we?
Yeah but with a sample size of 1, what useful conclusions can you draw from that ?
You can infer that humans are likely average for higher intelligences, because if you found only one example of something it's more likely to be an average example than an outlier.

>as the laws that govern organization of physical matter are universal, it follows that any other mind MUST be fundamentally similar to the human mind
There could be multiple "mind designs" that work for navigating reality. How would we know that there can only be one "design" that works well.

>I think one of those fundamental requirements is the use of language all languages have common universal properties, so if there is a language, there is a possibility of communication
This is more or less Chomsky's argument for universal grammar.
But I don't know enough linguistics.

>>11903
Spoiler alert. Not even linguists take Chomsky seriously

>>11900
>lol, individual or small groups of capitalists will get jail treatment from the state if they decide to break one of the fundamental pillars of capitalism - intellectual property enforcement

"intellectual property" isn't a type of property.
It's a distribution monopoly that is granted by the state. And that isn't just me doing ideological struggle, that's how it's defined in bourgeois law. Even the mythology for the origin story in the ruling narrative says that 2500 years ago some cooks from Sybaris (southern Italy) got a monopoly on their recipes. (Tho the real origin is in the late 18 century British empire, where somebody filed a trademark for a triangle https://trademarks.ipo.gov.uk/ipo-tmcase/page/Results/1/UK00000000001?legacySearch=False) I'm not shitting on Trademarks, those are reasonable for the most part

Intellectual "property" monopolies are not fundamental to capitalism.
Capitalism ignores it all the time, the biggest capitalists are not upholding it, "IP violations" as it's called in their technical jargon are rampant in every sector of the economy. The material reality is that they just treat it as a legal weapon to fuck with the competition, there even is a word for that "law-fare".

You can have capitalism where the state doesn't grant distribution monopolies and only upholds trademarks and in a limited sense, licenses. If you go to Asia most people don't really have a concept for it, with the exception of Japan, so it might be a western particularity.

If they persecute people in the west for publishing machine-learning enhanced binary reverse-engineering tools, it will be developed somewhere else. Big corporations probably are already pursuing this for the purpose of commercial/industrial espionage. If somebody develops this tool it'll get out somehow, it always does. Not allowing this to exist legally might also mean loosing out on a valuable tool for analyzing software security.

>Ghidra and NSA is a truly bizarre case to me
It might make sense for them to publish security enhancing tools like this because, they want computer systems to have a level of security that only they can break.

>>11902
>No-one knows if AI is even possible,
I'm a hard materialist, so I think that the human brain is doing information processing, using electrical signals and neuro chemistry.
If the brain is doing something special we don't know about, that means more discoveries for information theory.

I think that if we wanted to build computers that worked like a human brain, we have to build it out of mem-ristors, not tran-sistors.
Think of a mem-ristor as calculating and storing information at the same time. And on a physical level it works by manipulating ion-channels, which is qualitatively similar to neuronal ion-channels in the brain. If this interests you the buzzword to look for is "neuromorphic systems".
Theoretically it should be possible to have currently typical computer architecture run mind-emulation software, because von Neumann machines are universal computers. However it's not practical, it would have an enormous size, power-bill and waste-heat affecting local weather patterns.

>machine learning,. Yes. But that isn't AI, it's just a superficial imitation.
I guess machine learning is just one part of an artificial mind.
Star trek predicted that autonomous artificial minds would be achieved roughly 500 years from now, i think that's a reasonable time-table.

It's unclear to me what you mean with AI being a superficial imitation.
What makes it superficial ?

>>11904
That's saying something given their standards.

>>11906
>I'm a hard materialist, so I think that the human brain is doing information processing, using electrical signals and neuro chemistry.

If anything this makes you a vulgar materialist. Define information for me.

>>11908
>If anything this makes you a vulgar materialist
You are using this as a taunt in tribal discourse, that makes it meaningless as a theoretical criticism.

>Define information for me.
That's a huge subject that defies the requirements of brevity for posts such as this one, you have to read Claude E Shannon The Mathematical Theory of Communication. I tried to include the Book in the attachment but i can't upload files with a djvu file-extension, and after converting it to pdf the file size was too big. Sorry you have to get it from libgen or something in case you're interested

Remember deep fakes from a few years ago ?
That tech has gotten really good and it's now being used for video filters that make people look way better than they actually look

check this out

https://nitter.net/memotv/status/1629905913069879296#m

this is sort of AI related, and i didn't feel like making a new thread, so