221–230 of 238 posts
Earlier quoted context omitted.
I'm sorry if anything I said insulted you or seemed to be a comment on you personally. That wasn't my intention.
Apology accepted and please accept mine in return, that was much too spiky a reply from me.
Earlier quoted context omitted.
Not after RLHF tuning, due to the 'flattened logits' phenomenon (which is the logit-level version of the mode collapse OP documents at higher levels). All the temperature settings wind up yielding pretty much the same output, until you ramp it up so high that it falls apart completely. Completely unlike the base models where you can productively tune the temperature or use very high temperatures with some screening.
Hmm, it's hard to check without access to the prompts used in the paper, but I'm skeptical that the distributions seen in e.g. Figure 2 are so different that you would have crank up the temperature very much to bridge the gap. It looks to me like the entries that are 1-in-100 in the base model are just falling off the top-p cliff and getting set to 0.
So you don't know how any sampling would affect that. There could be only a few options at each token, which give rise to that, and higher temperature sampling may shift that around, but it doesn't ever restore the original base model behavior or restore all of the names erased by mode collapse. (Remember, the LLM is an agent, and when you are sampling, it is on-policy because you are letting it make choices of tokens, and it is steering the completion as a whole back to where it wants to be. With mode collapse, all roads lead to Rome, whether you like it or not.)
People do observe that increasing the temperature does not help, eg. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC10936766/ finds basically no difference going from 0 to 0.9 (!): https://www.ncbi.nlm.nih.gov/pmc/articles/PMC10936766/bin/po... Just the flattened logits (https://arxiv.org/pdf/2303.08774#page=12&org=openai) at work.
"Bias" implies the possibility of "unbiased language model" which seems to be in the category of things that are on one hand, COMPLETELY IMPOSSIBLE, and on the other, still likely to be sold on the market because market wants it so much?
Even assuming we can make an unbiased model (assuming by unbiased we mean something like "has a world model and reasoning that has no systematic deviation from reality"), we couldn't recognize the model as unbiased. I'd even wager that outside of research such a model would be completely unusable for practical applications. Both as individual humans and as collective societies we have a lot of biases. And judging by…
Yeah that’s a way’s off. An LLM is just a reflection of the text that humans write, and humans seem very far off from having world models and reasoning that accurately reflect reality. We can’t even reason about what the real differences are between men and women (plus countless other issues) because our pictures of reality are so warped by ‘points of view’.
(not quite wokism)
Earlier quoted context omitted.
> The nerve of me, to expand on my views as a discussion develops. Nothing wrong with expanding your views. But you've neither defended nor retracted your original argument. I'm trying to stick to that. > Of course you have lots of great points to make, but you can't share them with the likes of me. I don't have anything to say about your new argument (which may be great and compelling), I haven't thought through it…
Let's just be perfectly clear here. You asked: why should we treat these two cases differently? I answered: one of them involves harm and one can't possibly. You then refused to discuss the topic further. I never changed my mind. I never contradicted myself. I have nothing to retract. I simply elaborated on my views in response to your question. Regardless of whether you "didn't intend to introduce something new" - y…
I wasn't asking why we care about the treatment of humans. I was pointing out that your original argument makes no sense, because of what it would mean if you applied it to humans. You then responded with a different argument about humans and AIs being different, which - regardless of its merits in its own right - in no way relates to your original one and does nothing to rescue it.
Earlier quoted context omitted.
Let's just be perfectly clear here. You asked: why should we treat these two cases differently? I answered: one of them involves harm and one can't possibly. You then refused to discuss the topic further. I never changed my mind. I never contradicted myself. I have nothing to retract. I simply elaborated on my views in response to your question. Regardless of whether you "didn't intend to introduce something new" - y…
> You asked: why should we treat these two cases differently? I wasn't asking why we care about the treatment of humans. I was pointing out that your original argument makes no sense, because of what it would mean if you applied it to humans. You then responded with a different argument about humans and AIs being different, which - regardless of its merits in its own right - in no way relates to your original one and…
And I explained why this equivalence doesn't hold in my view. You have made no attempt to justify it. So why should I buy it? Are you obliged to buy every proposition I put forward? Obviously not, or we wouldn't be having this discussion.
> [It] in no way relates to your original [argument] and does nothing to rescue it.
My argument is only endangered if the equivalence holds - and again, I haven't bought it. You seem to feel entitled to my adopting your perspective, and to respond assuming your equivalence is valid. But you have to earn that by putting forward a good argument.
Instead of trying to convince me or explore our disagreement, you declare that I violated unstated terms and that the discussion cannot proceed. This is nonsensical.
Since it's clear you aren't willing to give me a fair hearing, I encourage you to show this conversation to someone you respect and trust to be honest with you, so that they can explain this to you in a way that will register.
Earlier quoted context omitted.
> You asked: why should we treat these two cases differently? I wasn't asking why we care about the treatment of humans. I was pointing out that your original argument makes no sense, because of what it would mean if you applied it to humans. You then responded with a different argument about humans and AIs being different, which - regardless of its merits in its own right - in no way relates to your original one and…
> I was pointing out that your original argument makes no sense, because of what it would mean if you applied it to humans. And I explained why this equivalence doesn't hold in my view. You have made no attempt to justify it. So why should I buy it? Are you obliged to buy every proposition I put forward? Obviously not, or we wouldn't be having this discussion. > [It] in no way relates to your original [argument] and…
No you didn't.
> Instead of trying to convince me or explore our disagreement, you declare that I violated unstated terms and that the discussion cannot proceed.
No I didn't.
> Since it's clear you aren't willing to give me a fair hearing, I encourage you to show this conversation to someone you respect and trust to be honest with you, so that they can explain this to you in a way that will register.
Right back at you.
Earlier quoted context omitted.
> I was pointing out that your original argument makes no sense, because of what it would mean if you applied it to humans. And I explained why this equivalence doesn't hold in my view. You have made no attempt to justify it. So why should I buy it? Are you obliged to buy every proposition I put forward? Obviously not, or we wouldn't be having this discussion. > [It] in no way relates to your original [argument] and…
> And I explained why this equivalence doesn't hold in my view. No you didn't. > Instead of trying to convince me or explore our disagreement, you declare that I violated unstated terms and that the discussion cannot proceed. No I didn't. > Since it's clear you aren't willing to give me a fair hearing, I encourage you to show this conversation to someone you respect and trust to be honest with you, so that they can e…
https://news.ycombinator.com/item?id=40703486
If I was unclear - the equivalence doesn't hold because one of these are moral entities and the other isn't.
> No I didn't.
https://news.ycombinator.com/item?id=40703471
https://news.ycombinator.com/item?id=40711859
> Right back at you.
Fair enough. I was already planning on it, so I'm happy to.
Earlier quoted context omitted.
"one art work from every decade of the last 500 years that would likely be highly diverse in style and content" It still might not be especially diverse if all 50 examples were from western European art. 500 years only takes us back to 1524 - not especially long and mostly from the same early modern period starting with the fall of Constantinople, the end of the Crusades, and the start of the Renaissance. I wouldn't…
> I wouldn't be surprised if 80% or more of the works ended up being some depiction of aspects of Christianity painted by a white male. Are you saying diversity in art is signified by the artist's race and sex?
