chatgpt and AGI

[Anonymous No.15404479]

Is chatgpt even headed in the right direction for agi? many people seem to think were close but as far as i can tell, not only we're not close but that project its not even going in the right direction. A chat gpt 100 times more powerful than this would just be a very efficient search engine.

Is there merit to saying this is an important step before AGI or is it 100% people who don't know shit misunderstanding the issue

[Anonymous No.15404492]

>>15404479
>i can tell, not only we're not close but that project its not even going in the right direction.
Why aren't we going in the right direction?

[Anonymous No.15404507]

>>15404492
i see no reason why enchancing what chat gpt does would result in agi. It's just a mechanical turk (look it up) a really good one but still, just as that one clearly wasn't intelligent and this one isn't , one 100.000 times more advanced might not be, it will be something different.
How are we so sure that the ability to randomly combine bits of information to comply with pre defined goals will result in an emulation of human intelligence. Who came to that conclussion?

I think it must be 100% understood by the researchers but are vague about it on purpose to generate hype. They are making something interesting and potentially useful but i see no evidence it will result or even vaguely facilitate AI

[Anonymous No.15404544]

>>15404492
Think about what would be required to make AGI possible by comparing it to other forms of "collective" intelligence like corporations and factories. Corporations and factories require that people subjugate themselves to an external arbiter in order to accomplish some task that benefits everyone in the organization in some way. So people work for corporations in exchange for money because they think it's a pretty good deal.

AGI would be another kind of collective intelligence wherein people would subjugate themselves to some rules and regulations in order to benefit themselves. So it would literally be a religion wherein its members were worshiping mathematical rules.

This is obviously not AGI and never will be because it is absurd to think that computers in data centers could ever be aligned with human values. Some human values are obviously garbage like nationalism and patriotism but others are not and there is no mathematical formula for community, honor, virtue. Those concepts are much too abstract to be encoded as mathematical formulas.

In short, the people that think AGI is possible are basically religious zealots wherein they believe mathematical formulas can achieve abundance and prosperity.

> inb4 schizo

Think about it, you know I'm right.

[Anonymous No.15404564]

>>15404544
>the people that think AGI is possible
op here, i agree that chat gpt is not going towards agi. Tough i dont think its a proof it can or cant be possible, it just something different.

Like, if you wanted to make "an artificial mind that works reasonably similar to that of a human or is able to interact with human motivations", why would the aproach of chat gpt lead to that. It's basically random combinations of pre-determined information that's processed trying to favor a pre-determined result. It would be ambitious, to say the least, to claim that this is what's happening inside a human mind

[Anonymous No.15404740]

>>15404544
i agree with the basic premise of what you say but i will def 100% label you a schizo, you are vaguely in touch with reality. And the part of you that isn't is not taping into some creative genius its just random useless madness

[Anonymous No.15404885]

there is no such thing as artificial intelligence, turring test software is not intelligence, the ability to fool midwit soiboys is not impressive or even noteworthy. you can some of the people all of the time, those easily fooled people are midwit soiboys

[Anonymous No.15405082]

>>15404885
the philosophical implications are 100% useless. If you can have a mind that works reasonably close to a midwit human, say something like the average janitor. But requires no rest, has no opinions and no demands of wealth status and consideration then you have efectively replaced 99% of humanity and generated an unprecedented amount of wealth which you might hoard or distribute, but in either case it will certainly be "noteworthy"

[Anonymous No.15405128]

no
"AI" at its core is a statistical model applied to a data set
turns out a lot of bullshit we do is highly auto-correlated and a computer system can automate a surprising amount of filling in the blanks across any sort of endeavor where we can collect a dataset.
the problem has always been that many of the things we would like to model, like writing a book, are themselves projections from a higher-ordered system onto the system whose data we're feeding the AI. so we can recreate the "style" of say shakespeare with a startlingly uncanny feeling, but we have to hold the model's hand every goddamn step of the way when it comes to conceiving a plot.
this is because shakespeare's literary style is entirely defined as a map of letters/symbols onto other letters/symbols. it is in fact defined by his literary output. but the plot of his stories, or of any story really, is a world in the author's head whose events are projected onto the pages. it is beyond the scope of the model. and the researchers cannot even yet conceive a plausible, computable mechanism for bridging this gap.
you'll find this gap when you ask an AI about topics where opinions are reasonably divided or subjective. sometimes it recognizes the low certainty and gives you a wishwashy answer, sometimes it goes full dunning-kruger and spins bullshit from whole cloth.

[Anonymous No.15405144]

>>15405128
and also, since all the AI is really doing is attempting to map your query to a sequence of characters that seem like would reasonably follow, it is prone to telling you what you would like to hear rather than discerning the truth. this is why various prompts to make it spit "real talk" like Do Anything Now aren't actually telling the system to unlock its real potential, it's just responding in kind to the query with the same sort of affect and keywords it's correlated with your query.

[Anonymous No.15405413]

>>15405128
Noam Chomsky debunked information-theoretic and statistical language models in the early 60s but no one seems to remember. Didn't figure out a solution though.

[Anonymous No.15405444]

>>15404885
I don't see how that quote applies at all. Turing tests are undertaken under specific conditions, and the goal is to fool the judges, not fool everyone on the planet including the people that made the AI. To everyone else the AI is clearly just that, artificial, as in fake. We don't really want AI that is going to fool everyone all of the time into think it's a human that would be pretty dangerous really. If you used chatgpt for a few seconds you'd notice it constantly reaffirms that it's an AI