I think the camera analogy is much more apt than the engine analogy, so this article was intriguing, but unfortunately the author falls into the rhetoric they decry when they complain about equating black holes with God, which is to just pull out a flourish like "monkeys at typewriters all the way down" to casually step over the difference between intelligence and mere computation, or to equivocate what is observed in neural networks with "intelligence" at all, and then by way of disclaimer to then dismiss intelligence completely. It would be much more interesting to discuss how machine learning gives us a lens into latent structures in data, and where that could be applied in the future, or even what that learning says about what we don't know about intelligence. And I was hoping that's where this would go, but the article seems to be attempting some grand theory of everything that's unlikely to have a practical application, which makes it difficult to discuss without first blocking out the millenarian aspects.
"This is what the stochastic-parrot takes (and the older monkeys-at-typewriters take) entirely miss."
I do hope people realize that the "stochastic parrots" thing is not a "take" but a 14 page research paper by four qualified authors, with a 158 item references section.
Not to mention we're not just stochastic parrots and monkeys on typewriters. We have reasoning. They can try to emulate it by putting ever more parameters but the reason we have nonsensical things like hallucinations is because they can't filter the output through reason before it comes out. Look at the lawyers getting caught with their pants down on ChatGPT. It's not actually reasoning a citation, it's just bullshitting what a citation might look like because a legal brief de novo needs to come from a place of reason.
The imitation will get closer and closer to reasoning and some will mistake it for superior reasoning. I don’t think it’s impossible for computers to reason in analogous ways to us but the current road alone is not it, in my opinion. I think for artificial intelligence to florish it should have some form of artificial life to it, some system with a bit more complexity and feedback loops than anything that I know of today.
> I do hope people realize that the "stochastic parrots" thing is not a "take" but a 14 page research paper by four qualified authors, with a 158 item references section.
A whale "is" a mammal, as "is" a mouse. But a whale is not "equal to" a mouse.
> by four qualified authors
What is the (comprehensive) meaning of this symbol: "qualified"?
> What is the (comprehensive) meaning of this symbol: "qualified"?
Why the obtuse questioning? You can look up the authors and their qualifications are plain to see. Look, here's the first author, you can see for yourself.
Why the dodging of the question? Might you be leveraging culture? At your day job do you use this same technique against computers, shaming compile errors into compliance with your desires?
You seem like an more intelligent than average human, why not demonstrate it rather than merely imply it? No need to constrain yourself to cultural norms of politeness and meme magic, feel free to blow my doors off.
I asked a pointed question - do you have the ability to answer it? Or even better: do you have the ability to even try, in public? Internet forum discussions are serious business, so I can appreciate both aversion, and silence[1].
>> What is the (comprehensive) meaning of this symbol: "qualified"?
Do you know the(?) answer to this question?
Noteworthy: you are in no way obligated to answer to me, though there may be causal ramifications of you are unable to.
Oh my, look at all the culturally unusual fodder I have provided for you to leverage cultural norms against! I did my best to make it irresistible.
Have you any more memes for the class, intelligent Western human? I am here to learn, and learn from you I intend to do.
[1] Here is an excellent opportunity to proclaim (sincerely!) confusion due to a non-normative usage of language and dismiss me as crazy/etc. Not that it's necessary in the first place - we are on your, and Dang's, and YCombinator's turf - you can literally declare victory at any time, and win! Do you know why this is?
META: dang, if I was you I would would reduce the allowable time limit for edits, at least on a per user basis as you (your company) has done with posts. As reality becomes more complex, it is rational to increase your (platform's) ability to distort it - to me, this is fair game play...though one wouldn't want to reveal it is a game, of course. Just an idea.
> I asked a pointed question - do you have the ability to answer it?
I am refusing to accept your framing (being that we should discuss qualification in detail) as I suspect you do not call in to question the qualifications of the authors (or the meaning of the word qualifications) when an AI paper being discussed has been written by men, rather than women.
>>> What is the (comprehensive) meaning of this symbol: "qualified"?
> Do you know the(?) answer to this question?
Comprehensive meaning? Nope. But I don't accept that I must be able to comprehensively articulate the meaning of every word I use. I accept that the authors of the paper have enough education and experience to write a paper on this subject that is worth examination.
The biggest reason I dismissed you originally is that your style of conversation is coming off to me more argumentative than anything else, and it is my experience that arguing with people on the internet is a waste of time. Plenty of people on this site treat others with respect, and reasonable people can disagree about a great many things, and so I am willing to explain my reasoning to them in detail. But when someone comes in with what seems like an argumentative state of mind, I either ignore them or make a comment that purposefully ignores their framing of the question, knowing that is a trap.
> Oh my, look at all the culturally unusual fodder I have provided for you to leverage cultural norms against! I did my best to make it irresistible.
This is a great example of argumentative style that I personally will just ignore.
> we are on your, and Dang's, and YCombinator's turf - you can literally declare victory at any time, and win!
More evidence, in my opinion, that you are viewing this not as a respectful conversation between reasonable people, but some weird kind of internet battle field. I literally don't give a shit about winning, so your assumptions about my motivations are flawed and this explains why you find my response confusing.
> I don't accept that I must be able to comprehensively articulate the meaning of every word I use.
Did I claim that you are obligated to?
You certainly don't need to be able to. Worse, you may not even be able to see the benefit of a species using language that is highly and consistently understandable (as is common in science, engineering, law, etc).
Let's hope there aren't any negative consequences from this ideology, which seems to be a cultural norms, supported by the vast majority of people across most any ideology ("pendants" being an exception - for fun, look up the meaning of that word in the dictionary).
Say, what's your stance on climate change, and the various wars The Humans have going on?
In all seriousness, that take was polemic and arguably bad-faith when it was published; and it is much more egregiously so now.
The good news is we're going to hear much, much less about "stochastic parrots" when multi-modal models come into their own. The "faked" Gemini video may have been faked but it won't be fake for particularly long, that should be obvious to anyone in or closely following the field.
The premise that human beings do something substantially different from correlate multilevel abstractions is looking pretty weak.
If one prefers to take the other side as primary and start arguing that what they mean by parrots applies about as much to humans, well, then we can start using that language.
This is some pretty poor philosophy... or philosophy of science. This kind of seems like the common mistake where people confuse the stories about science as the science itself. It seems to be fitting a lot of metaphors together as its evidence.
This feels like calling every piece of literary work, be it a comment, or a novel as a part of the Library of Babel, instead of one's genuine creative endeavour. While that camera analogy felt like an eye-opener at once, it feels like an attempt to trivialize; we are not inventing anything new, but discovering what already exists. I disagree.
Has colors of the free will debate. We may not actually be creative, but that doesn't stop the illusion of creativity from being any less delicious or meaningful!
18 comments
[ 3.8 ms ] story [ 48.5 ms ] threadI do hope people realize that the "stochastic parrots" thing is not a "take" but a 14 page research paper by four qualified authors, with a 158 item references section.
https://dl.acm.org/doi/epdf/10.1145/3442188.3445922
A whale "is" a mammal, as "is" a mouse. But a whale is not "equal to" a mouse.
> by four qualified authors
What is the (comprehensive) meaning of this symbol: "qualified"?
Why the obtuse questioning? You can look up the authors and their qualifications are plain to see. Look, here's the first author, you can see for yourself.
https://faculty.washington.edu/ebender/
Why the dodging of the question? Might you be leveraging culture? At your day job do you use this same technique against computers, shaming compile errors into compliance with your desires?
You seem like an more intelligent than average human, why not demonstrate it rather than merely imply it? No need to constrain yourself to cultural norms of politeness and meme magic, feel free to blow my doors off.
I asked a pointed question - do you have the ability to answer it? Or even better: do you have the ability to even try, in public? Internet forum discussions are serious business, so I can appreciate both aversion, and silence[1].
>> What is the (comprehensive) meaning of this symbol: "qualified"?
Do you know the(?) answer to this question?
Noteworthy: you are in no way obligated to answer to me, though there may be causal ramifications of you are unable to.
Oh my, look at all the culturally unusual fodder I have provided for you to leverage cultural norms against! I did my best to make it irresistible.
Have you any more memes for the class, intelligent Western human? I am here to learn, and learn from you I intend to do.
[1] Here is an excellent opportunity to proclaim (sincerely!) confusion due to a non-normative usage of language and dismiss me as crazy/etc. Not that it's necessary in the first place - we are on your, and Dang's, and YCombinator's turf - you can literally declare victory at any time, and win! Do you know why this is?
META: dang, if I was you I would would reduce the allowable time limit for edits, at least on a per user basis as you (your company) has done with posts. As reality becomes more complex, it is rational to increase your (platform's) ability to distort it - to me, this is fair game play...though one wouldn't want to reveal it is a game, of course. Just an idea.
I am refusing to accept your framing (being that we should discuss qualification in detail) as I suspect you do not call in to question the qualifications of the authors (or the meaning of the word qualifications) when an AI paper being discussed has been written by men, rather than women.
>>> What is the (comprehensive) meaning of this symbol: "qualified"? > Do you know the(?) answer to this question?
Comprehensive meaning? Nope. But I don't accept that I must be able to comprehensively articulate the meaning of every word I use. I accept that the authors of the paper have enough education and experience to write a paper on this subject that is worth examination.
The biggest reason I dismissed you originally is that your style of conversation is coming off to me more argumentative than anything else, and it is my experience that arguing with people on the internet is a waste of time. Plenty of people on this site treat others with respect, and reasonable people can disagree about a great many things, and so I am willing to explain my reasoning to them in detail. But when someone comes in with what seems like an argumentative state of mind, I either ignore them or make a comment that purposefully ignores their framing of the question, knowing that is a trap.
> Oh my, look at all the culturally unusual fodder I have provided for you to leverage cultural norms against! I did my best to make it irresistible.
This is a great example of argumentative style that I personally will just ignore.
> we are on your, and Dang's, and YCombinator's turf - you can literally declare victory at any time, and win!
More evidence, in my opinion, that you are viewing this not as a respectful conversation between reasonable people, but some weird kind of internet battle field. I literally don't give a shit about winning, so your assumptions about my motivations are flawed and this explains why you find my response confusing.
Did I claim that you are obligated to?
You certainly don't need to be able to. Worse, you may not even be able to see the benefit of a species using language that is highly and consistently understandable (as is common in science, engineering, law, etc).
Let's hope there aren't any negative consequences from this ideology, which seems to be a cultural norms, supported by the vast majority of people across most any ideology ("pendants" being an exception - for fun, look up the meaning of that word in the dictionary).
Say, what's your stance on climate change, and the various wars The Humans have going on?
In all seriousness, that take was polemic and arguably bad-faith when it was published; and it is much more egregiously so now.
The good news is we're going to hear much, much less about "stochastic parrots" when multi-modal models come into their own. The "faked" Gemini video may have been faked but it won't be fake for particularly long, that should be obvious to anyone in or closely following the field.
The premise that human beings do something substantially different from correlate multilevel abstractions is looking pretty weak.
If one prefers to take the other side as primary and start arguing that what they mean by parrots applies about as much to humans, well, then we can start using that language.
Otherwise it's poppycock.
Did you spend any time/compute in the massive hyperlinked space underlying this part?
1. Whatever the hell Leibniz was thinking about:
https://en.wikipedia.org/wiki/Calculus_ratiocinator
a) If so, how confident are you that you understood it, on a scale of 1 to 10?
b) If you did not, why did you not?
c) Do you know what Leibniz was talking about? Because it may be a good question. (Or, it may not be.)
> This kind of seems like...
Speaking of philosophy: have you any interest in what it is?
> where people confuse
People are indeed silly. One problem with silliness is, it tends to render the patient unable to accurately self-diagnose.