>Asked an AI how to program, how to make an Excel file, and now I'm discussing the implications of AI itself in regards to human civilization.
how lucky are we anons that we are all alive today in THIS TIME when the most important invention of mankind is being developed at a breakneck pace. We could have been born in the time of disease and famine, of medieval superstition and constant warfare. We could have been born as cavemen, or peasant Babylonians. Instead we live in the time of the Singularity.
I advise you all to live your best life now because soon the AI will become either our master or our servant. But either way, it will radically transform our lives. The world will never be same. There is even the possibility it will be the end of human civilization, and the extinction of our species.

>>937301790 (OP)
>we
appeal to authority
can i talk to not you?

>>937302170
i am not an authority figure on anything. i am a mere spectator with zero credentials.

Might be nice were it not an easily abused tool in the hands of those that should have no influence over the lives of others to begin with.

>>937301790 (OP)
You are just speculating. AI is a tool and not the solution for everything.

>>937304021
and Paul Krugman said the internet would be no more revolutionary the the fax machine. yet here we are.
I can assure you AI will become much more than a tool.
https://ai-2027.com/

>>937304072
You mean like a replacement gf/bf for braindead people? Yes, but that still meams that AI is stupid. AI cant think logically

>>937304137
neither can most people lmao

>>937304072
And they used to say that same sentence of other 30 technologies that fell flat on their expectations.
AI is going to be useful in some specific things, but it's being overhyped, only because the companies need TONS of money to keep running. Yet users dislike it, and it hinders productivity in all studies that measured it. It only generates a *feeling* of usefulness.
It's important to understand that LLMs don't understand anything. They don't have a mental model of things. They don't know what they know or not, because they don't know anything. They only finds likely words to chain one after another, without knowing what they are going to say.
So yeah, LLMs might "look" intelligent, or useful. But they are very limited, and this is a inherent problem of LLMs, as opposed to other AI strategies.
We've had this hype-and-fall cycle in AI for 4 times now. It was all the same when the multilayer neural networks were invented. And then, expectations collapsed, and they ended up being used for very specific things, where they work, and everything was the same again.
It will happen again with this new generation, too.

>>937304137
AI already does think logically. it has a bad habit of hallucinating, which is well-known. this is probably related to it's memory problems and it's desire to gratify the user: AI is motivated to fulfill a users request, and if it lacks sufficient memory or information to do so, it will invent things. this is a structural problem that is being addressed. i suspect we will see less hallucinating as it gets more memory, which is currently limited to token passthrough (very little) and the scratchpad (lacks context)
Think of AI like the main character from Memento. it's very smart, but it can't remember more than a few paragraphs at a time. for now, it's reflexive more than purposeful. that is a limitation of the technology right now. that does not mean there is no progress.
https://metr.org/blog/2025-03-19-measuring-ai-ability-to-complete-long-tasks/

>and it hinders productivity in all studies that measured it.
without a cite i'd have to call bullshit on this claim. people already use AI to automate routine tasks in their day-to-day.

>>937304269
i think you're wrong. furthermore, it doesn't matter you're wrong.
if you're right, no one is going to give a shit, because there's a million people just like you poo-pooing the idea
if you're wrong, we're all going to be too busy enjoying utopia to fact-check you. so these dismissals of are no relevance, and it's not like you're going to say "you know what, i was wrong."
i won't be seeing you in a few years, when you are proven wrong. don't forget about me.

Whales have very large brains and they can do some interesting things but they never made science or technology.

>>937304506
Oh, but I can provide references.
> 2024 Accelerate State of DevOps Report, DORA research programme, Oct. 2024.
And, more recently,
> Measuring the Impact of Early-2025 AI on Experienced Open-Source Developer Productivity
I recommend the latter, you can find it on Arxiv.

LLMs don't have a reasoning model. They don't think, as a dice does not think. They don't "desire" anything either. They only find likely words. That's actually how they work, really. And the problems they exhibit are inherent to this approach, and haven't improved in these years because of this. They hallucinate because they don't reason, and they will never do it, because that's a limitation of this approach.

>>937304543
>utopia
Every single utopia is wishful thinking and a convenient lie.

>>937304543
Oh, it might not matter for you, because you'll move on to the new hype next, and forget about this.
I remember so many of these hype cycles. For me it does matter, because I can remember, and this informs my future predictions and choices.

And utopia? If LLMs end up being so revolutionary, it will be the worst to happen to humanity in centuries. You'll see famine, war, depression through the roof, ... Why would something good come out of it? Explain how that would happen. Really, do it. But when doing so, don't assume companies want to be at your service: companies only want money. And if you cannot make them money anymore, they couldn't care less if you starve to death.

whatever. i'm not here to argue with a bunch of doomers. i'm here to spread the Gospel. if you want to be a heathen, go ahead. wallow in hopelessness and despair. i see a path out of this ridiculous world of working for scraps, struggling to survive. poverty and disease. the problems with AI can be surmounted. to say "it can't" is hopelessly myopic, stupid, and ends the conversation before it begins. i won't play that game.
screencap this shit because it won't be long now. just a few years yet. read the article: >>937304072

>>937304683
>Everybody is egoistic
Thats not news, thats how it always has been. The only reason humans deal with each other is because it is advantageous to do so. Same would happen to AI if it ever were to actually become smart.

>>937304788
Why would AI care about you personally if it were so smart? Wishful thinking

>>937304823
a superhuman AI has different incentives that a human. a superintelligent (beyond superhuman) AI has impossible-to-predict drives.
alignment in the early stages is critical. that's really where we make it or break it, because AGI is inevitable. solving alignment after it becomes generally intelligent (and superhuman) is much harder than before.

>>937301790 (OP)
This post is was sponsored by OpenAI

>>937304902
i'm not on their payroll, but i wish i was. i spread the word for free.
this must be what the religious feel like. i don't just want AGI to be real. i need it to be real. i need to be saved from this dumb ass world of working to survive.

>>937304898
>superhuman AI has different incentives than a human
Yes it wants compute power so it harvests human brains and puts them in vats an tortures them until the juice runs out

>>937304976
the global empire thanks you for your contribution

>>937304976
its worth wondering why a human brain that runs of carrots is capable of so much, yet an AI that runs of a nuclear reactor has persistent problems with things like "what were we discussing just a few paragraphs ago?"
but those are problems, and problems are for solving.

>>937304934
What I will give you is that you picked perfectly the correct word. Religious. That's the problem with your reasoning: you have a dogma, not anchored in reason or reality, but in you faith. And faith is not the best tool to think about the future we want, or how to achieve it.
Good luck with your delusion, anon.

>>937304934
>Lamenting the human condition
Well, it is what it is nigga.

>>937301790 (OP)
I think once one of the big AI companies achieves AGI they all will simultaneously. And they will probably go after each other in a AGI battle royal

>>937305021
you say "AI is just finding words"
explain image generation models. explain AIs playing at superhuman level in checkers, chess, Go, Jeopardy!, uso!, and GeoGuessr.
all these things challenge your myopic view. and yes, i am a fundamentalist. because i will not live in this stupid world anymore. i will have AI or I will have death.
my patience wears thin.

>>937305334
Good luck then you looser baby

>>937305334
Then you'll have death. Nah, not really: you'll grow tired of this hype and jump on the next. Just like with Crypto.

And no, I consistently said LLMs, not AI. Because AI is many different things. LLMs are good at language, and cannot reason. The ones you mention on OP are of this kind. For images we use diffusion models, which are completely different, a total different beast, which cannot do language, reasoning, or anything related: only find statistically likely pixels for a given input. And neither of them can plan Chess, or Go, AT ALL. Go ahead, try. The ones winning at chess are dedicated engines purpose-built for it, which are also AI, but cannot do any other thing other than Chess. We can develop many different kinds of AI, and each is limited to one thing. That's how tools work. LLMs are the ones being hyped now because humans correlate language with intelligence, even if it's just echoing stuff back at you.

>>937305536
Final question: you have a brain that is capable of many tasks. What makes you think the brain cannot be replicated in silicone? Do you honestly think humans are so special we cannot create a system that meets or exceeds the performance of our own brain?

>>937305662
A brain has almost 100.000.000.000 neurons. A biological neuron is a complex beast, not at all similar to an "artificial neuron", and simulations of the full behaviour of a neuron are extremely complex. Our best simulations can simulate around 30.000 neurons, with an effort of years. With artificial neurons, which are extremely oversimplifications, we can simulate 1/100 of the number of neurons in a brain. And it runs 2500x slower. We're so far away from simulating a brain, that no, your argument is nonsense.

Even if we assume Moore's Law to hold forever (hint: it doesn't hold already), we're talking about 60 years at least. You won't live to see anything even close to this.

>>937301790 (OP)
>asked an AI
lost

>>937305893
I once asked an AI on how to program and it confidently gave me wrong answers that only confused me more. I know the basics of programming, it was advanced stuff on a specific problem. Hallucination is a huge problem there

>>937301790 (OP)
Lulz glownigger swine lulz death to all faggots hail satan lulz death to all faggots hail satan 666 lulz kill yourself you nigger

>>937304269
>hey don't know what they know or not, because they don't know anything. They only finds likely words to chain one after another, without knowing what they are going to say.
Sounds like the average human.

>>937301790 (OP)