Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

A non-anthropomorphized view of LLMs

addxorrol.blogspot.com

Post reply on HN
Print view View original

401–410 of 432 posts

Re: A non-anthropomorphized view of LLMs

#401

Post by Timwi »

tdullien wrote: post #264
Timwi wrote: post #161

The author seems to want to label any discourse as “anthropomorphizing”. The word “goal” stood out to me: the author wants us to assume that we're anthropomorphizing as soon as we even so much as use the word “goal”. A simple breadth-first search that evaluates all chess boards and legal moves, but stops when it finds a checkmate for white and outputs the full decision tree, has a “goal”. There is no anthropomorphizi…

Author here. I am entirely ok with using "goal" in the context of an RL algorithm. If you read my article carefully, you'll find that I object to the use of "goal" in the context of LLMs.

If you read the literature on AI safety carefully (which uses the word “goal”), you'll find they're not talking about LLMs either.

Re: A non-anthropomorphized view of LLMs

#402

Post by seyebermancer »

flir wrote: post #233
LeonardoTolstoy wrote: post #192

Earlier quoted context omitted.

What does a submarine do? Submarine? I suppose you "drive" a submarine which is getting to the idea: submarines don't swim because ultimately they are "driven"? I guess the issue is we don't make up a new word for what submarines do, we just don't use human words. I think the above poster gets a little distracted by suggesting the models are creative which itself is disputed. Perhaps a better term, like above, would…

I really like that, I think it has the right amount of distance. They don't write, they model writing. We're very used to "all models are wrong, some are useful", "the map is not the territory", etc.

What about they synthesize?

Ties in with creation from many and synthetic/artificial data. I usually prompt instruct my coding models more with “synthesize” than “generate”.

Re: A non-anthropomorphized view of LLMs

#403

Post by lowsong »

Al-Khwarizmi wrote: post #117

I have the technical knowledge to know how LLMs work, but I still find it pointless to not anthropomorphize, at least to an extent. The language of "generator that stochastically produces the next word" is just not very useful when you're talking about, e.g., an LLM that is answering complex world modeling questions or generating a creative story. It's at the wrong level of abstraction, just as if you were discussing…

> The language of "generator that stochastically produces the next word" is just not very useful when you're talking about, e.g., an LLM that is answering complex world modeling questions or generating a creative story.

But it isn't modelling. It's been shown time, and time, and time again that LLMs have no internal "model" or "view". This is exactly and precisely why you should not anthropomorphize.

And again, the output of an LLM is, by definition, not "creative". Your saying we should anthropomorphize these models when the examples you give are already doing that.

Re: A non-anthropomorphized view of LLMs

#404

Post by tdullien »

Timwi wrote: post #401
tdullien wrote: post #264

Earlier quoted context omitted.

Author here. I am entirely ok with using "goal" in the context of an RL algorithm. If you read my article carefully, you'll find that I object to the use of "goal" in the context of LLMs.

If you read the literature on AI safety carefully (which uses the word “goal”), you'll find they're not talking about LLMs either.

I think the Anthropic "omg blackmail" article clearly talks about both LLMs and their "goals".

Re: A non-anthropomorphized view of LLMs

#405

Post by andy12_ »

ACCount36 wrote: post #398
andy12_ wrote: post #282

Earlier quoted context omitted.

What? No. The intermediate hidden states are preserved from one token to another. A token that is 100k tokens into the future will be able to look into the information of the present token's hidden state through the attention mechanism. This is why the KV cache is so big.

KV cache is just that: a cache. The inference logic of an LLM remains the same. There is no difference in outcomes between recalculating everything and caching. The only difference is in the amount of memory and computation required to do it.

The same can be said about any recurrent network. To predict the token n+1 you could recalculate the hidden state up to token n, or reuse the hidden state of token n from the previous forward pass. The only difference is the amount of memory and computation.

The thing is that, fundamentally, an auto-regressive transformer is a model whose state grows linearly with each token without compression, which is what bestows them with (theoretical) perfect recall.

Re: A non-anthropomorphized view of LLMs

#406

Post by ben_w »

jibal wrote: post #289
CuriousSkeptic wrote: post #193

Earlier quoted context omitted.

> Why don't LLMs get frustrated with you if you ask them the same question repeatedly? To be fair, I have had a strong sense of Gemini in particular becoming a lot more frustrated with me than GPT or Claude. Yesterday I had it ensuring me that it was doing a great job, it was just me not understanding the challenge but it would break it down step by step just to make it obvious to me (only to repeat the same errors,…

Point out to an LLM that it has no mental states and thus isn't capable of being frustrated (or glad that your program works or hoping that it will, etc. ... I call them out whenever they ascribe emotions to themselves) and they will confirm that ... you can coax from them quite detailed explanations of why and how it's an illusion. Of course they will quickly revert to self-anthropomorphizing language, even after pr…

While they will agree with you that they don't, it's also the case that this is via training.

Consider that we have recordings of Brent Spiner covered in white paint and wearing yellow contact lenses claiming to have no emotions, not because he didn't, but because he was playing a role, which is also something we know LLMs can do.

So we don't know for sure if LLMs do or don't have qualia, irregardless of what they say, and won't until we have a more concrete idea of what the mechanism is behind that sense of the phrase "mental state" so we can test for their presence or absence.

Re: A non-anthropomorphized view of LLMs

#407

Post by Tijdreiziger »

scarface_74 wrote: post #381
Tijdreiziger wrote: post #320

Earlier quoted context omitted.

> I’m sure you knew that your code was running on computers somewhere even when you first started and wasn’t running in a literal “cloud”. Meh, I just knew that the browser would display HTML if I wrote it, and that uploading the HTML files made them available on my domain. I didn’t really think about where the files went, specifically. Try asking an average high school kid how cloud storage works. I doubt you’ll get…

If we can’t count on people on Hacker News to know that code runs on computers, what is this forum for?

Is Hacker News only for experienced coders who are already familiar with the inner workings of the ‘cloud’?

I’d like to think that this forum is also a place for the proverbial high school kid, who’s just learned JavaScript and deployed their first site to Vercel using their school Chromebook, to learn a thing or two from the greybeards.

Re: A non-anthropomorphized view of LLMs

#408

Post by Tijdreiziger »

godelski wrote: post #360
Tijdreiziger wrote: post #306

Earlier quoted context omitted.

Why would everyone know that? Not everyone has experience in sysops, especially not beginners. E.g. when I first started learning webdev, I didn’t think about ‘servers’. I just knew that if I uploaded my HTML/PHP files to my shared web host, then they appeared online. It was only much later that I realized that shared webhosting is ‘just’ an abstraction over Linux/Apache (after all, I first had to learn about those t…

I think they fumbled with wording but I interpreted them as meaning "audience of HN" and it seems they confirmed. We always are speaking to our audience, right? This is also what makes more general/open discussions difficult (e.g. talking on Twitter/Facebook/etc). That there are many ways to interpret anything depending on prior knowledge, cultural biases, etc. But I think it is fair that on HN we can make an assumpt…

See my reply here: https://news.ycombinator.com/item?id=44501642

Re: A non-anthropomorphized view of LLMs

#409

Post by lostmsu »

FeepingCreature wrote: post #211
latexr wrote: post #172

Earlier quoted context omitted.

> the same fallacy as claiming that crows can't be black because they have DNA of a bird. What fallacy is that? I’m a fan of logical fallacies and never heard that claim before nor am I finding any reference with a quick search.

(Not the parent) It doesn't have a name, but I have repeatedly noticed arguments of the form "X cannot have Y, because ". I wanna call it "fallacy of reduction" maybe: the idea that because a trait can be explained with a process, that this proves the trait absent. (Ie. in this case, "LLMs cannot think, because they just predict tokens." Yes, inasmuch as they think, they do so by predicting tokens. You have to actual…

It's much simpler than that. X is in B therefore X is not in A is what being said, and this statement simply doesn't make sense unless you have a separate proof that A and B don't intersect.

Re: A non-anthropomorphized view of LLMs

#410

Post by godelski »

Tijdreiziger wrote: post #408
godelski wrote: post #360

Earlier quoted context omitted.

I think they fumbled with wording but I interpreted them as meaning "audience of HN" and it seems they confirmed. We always are speaking to our audience, right? This is also what makes more general/open discussions difficult (e.g. talking on Twitter/Facebook/etc). That there are many ways to interpret anything depending on prior knowledge, cultural biases, etc. But I think it is fair that on HN we can make an assumpt…

See my reply here: https://news.ycombinator.com/item?id=44501642

I did. I referenced it too
Post reply on HN