133
submitted 2 months ago* (last edited 2 months ago) by Posadas@hexbear.net to c/chapotraphouse@hexbear.net
you are viewing a single comment's thread
view the rest of the comments
[-] Jabril@hexbear.net 31 points 2 months ago

Haven't seen it said yet but if the CPC can get a dialectical materialist AI up and running, and plug that baby into the internet, we might get the good singularity where we still get to exist as humans but in harmony with the ultimate central planning AI. If there is still a possibility for a Star Trek space communist future, something along these lines would probably be the only way. If the westerners don't want to have a revolution, the 5 Heads Bot can just shut off the stock market, all the missile silos, etc and end imperialism for us.

[-] Wheaties@hexbear.net 64 points 2 months ago

The programmes called "AI" are nothing like what we actually think when we hear "Artificial Intelligence".

You know how your phone's keyboard uses what you've already written to statistically predict the next most likely word you want to type? It's like that. The "AI" programme strings together sentences by selecting the most statistically likely word, one after another. The difference is, it's statistical model is based on a copy of everything ever written on the internet, rather than just what one person happens to type on their phone.

[-] semioticbreakdown@hexbear.net 38 points 2 months ago

Its in the name people... please... its literally just a model of language... a big-ass model of language....

[-] TheLastHero@hexbear.net 32 points 2 months ago

yes, please people stop falling for the marketing bullshit. There is nothing "intelligent" about these language models, it is not "thinking" you're just being impressed by their emulation abilities

[-] Jabril@hexbear.net 9 points 2 months ago* (last edited 2 months ago)

Of course LLMs aren't a simulation of consciousness with the same abilities of a human, the idea is that if a model was trained first on Marxist theory and history before taking in more information through that perspective there could be a point where this can be used to simulate economic models that would be useful for economic planning. It could be used to simulate contradictions and formulate strategies to navigate them in organizing spaces. It could be used for propaganda purposes, like if an average person asks it questions it would default to discussing the topic from a revolutionary angle. If some American goes on Deepseek to ask about how to convince their boss to give them a raise, Deepseek should default to teaching the person how to unionize their workplace instead of helping them form a good argument to convince the boss alone. There are a lot of use cases for an LLM trained this way, and this type of work would pave the way for greater advancements as the technology advances and we inevitably do get closer to a science fiction understanding of AI which is obviously not what a LLM is.

There are a lot of leftists here have a reactionary stance on this technology because of the way it is being used by capitalists, in the same way that anarchists have a reactionary stance on the state because of the way it is being used by capitalists. My wish casting fan fiction about a "good" AI existing one day in the future instead of the 99% more commonly thought idea that if an AI of this type would ever be developed it would kill us all is obviously bloomer cope. We'll be long dead before the technology gets there because the great minds of the left take a post about a marxist leninist dialectial materialist bot as a serious analysis of humanity's current technological progress and feel the need to critique it.

Edit: this post isn't directed towards the person I am responding to particularly, there has just been an ongoing undercurrent around this issue which is what I am speaking to more broadly

[-] Le_Wokisme@hexbear.net 8 points 2 months ago

Of course LLMs aren't a simulation of consciousness with the same abilities of a human, the idea is that if a model was trained first on Marxist theory

there could be a point where this can be used to simulate economic models t

but they don't simulate anything, they're word calculators

[-] Jabril@hexbear.net 2 points 2 months ago

How are you defining simulation? We can already generate images, videos, 3D models, text, interpret data and train models on particular data to base any of that generation on with currenly available platforms.

[-] Le_Wokisme@hexbear.net 9 points 2 months ago

dumping out scrabble tiles doesn't simulate systems.

[-] Jabril@hexbear.net 2 points 2 months ago

AI is already being used to assist simulation. One team used it to train robots by taking photos of a room and having the AI simulations train the robot on movements virtually instead of having to physically repeat the tasks in a real space. A quick search will yield many examples of the work being done that will allow the types of simulations you don't see now being done in 5-10 years.

[-] Le_Wokisme@hexbear.net 7 points 2 months ago

the AI simulations train the robot on movements virtually

that sounds like not an LLM

[-] Jabril@hexbear.net 1 points 2 months ago

I didn't say it was an LLM, other people brought up LLMs in response to my comment

[-] Wheaties@hexbear.net 3 points 2 months ago

I think computers themselves are already the under-utilized tool for economic planning and coordination. Perhaps LLMs (or some other sort of trained neural net) have a role in that, though I think the way it spits out answers without being able to go back and follow the steps it took to get there makes it a bit unreliable for economic modeling. Honestly, I've seen a pretty compelling case put forward that all we really need is some open-source algebra equations, and a dedicated network for incorporating worker feedback and real time data.

And, I'm... skeptical about the effectiveness of online messaging in general. It's good for getting ideas out there, but real organizing happens offline, between people. Ideally, we want them to be able to recognize, analyze, and work through contradictions themselves - rather than relying on the computer to hand them answers.

[-] Jabril@hexbear.net 1 points 2 months ago

I generally agree with all of this where we are now, especially the last part in regards to organizing.

I'm imagining that there are two concurrent timelines, one where machine learning and related technology continues to develop at an increasingly rapid pace, and another where westerners are under the heel of capitalism, increasingly desperate for change, but more or less alienated from revolutionary theory and practices due to their settler/colonial/fascist base ideologies preventing them from accepting the solutions to their problems.

Playing out that scenario, I could foresee a time where the technology (all machine learning, neural networks, "AI," and not LLMs particularly) has advanced to a point that it is more useful than the average American leftist in finding solutions to American problems, because American leftists are inhibited by the aforementioned ideologies and show no signs of letting them go. Many are doubling down these days.

Even now, the most talented organizers I know are mostly bogged down by the reproductive labor of keeping organizations afloat, and if even a third of that could be offloaded to machines and just touched up by the humans, it would save hundreds of hours a year that could go back into human to human interactions

load more comments (43 replies)
this post was submitted on 03 May 2025
133 points (100.0% liked)

chapotraphouse

13925 readers
646 users here now

Banned? DM Wmill to appeal.

No anti-nautilism posts. See: Eco-fascism Primer

Slop posts go in c/slop. Don't post low-hanging fruit here.

founded 4 years ago
MODERATORS