Now chart “I understand the issue now”
I'm absolutely right
151–160 of 277 posts
Re: I'm absolutely right
#152Earlier quoted context omitted.
Gay male marriages remarkably have the lower divorce rates than heterosexual marriages, while lesbian female marriages have higher divorce rates. Multiple studies show that lesbians consistently have far higher divorce rates than gays. This implies a level of neuroticism with females, that they probably do need to be appeased more, and that if you have two needy people who need to be appeased it's probably not going…
Funnily enough, male gay marriages also have by far the lowest domestic violence rate, and female gay marriages have the highest!
Reported rate. Gay men, like straight men, often underreport being victimized by their partners.
Re: I'm absolutely right
#153OK, so I love this, because we all recognize it. It's not fully just a tic of language, though. Responses that start off with "You're right!" are alignment mechanisms. The LLM, with its single-token prediction approach, follows up with a suggestion that much more closely follows the user's desires, instead of latching onto it's own previous approach. The other tic I love is "Actually, that's not right." That happens…
> The LLM, with its single-token prediction approach, follows up with a suggestion that much more closely follows the user's desires, instead of latching onto it's own previous approach. Maybe? How would we test that one way or the other? If there’s one thing I’ve learned in the last few years, it’s that reasoning from “well LLMs are based on next-token prediction, therefore ” is a trap. The relationship between the…
I don’t think there is any basis for GP’s hypothesis that this is related to the cursor being closer to the user’s example. The attention mechanism is position independent by default and actually has to have the token positions shoehorned in.
Re: I'm absolutely right
#154Earlier quoted context omitted.
Sorry if that felt dirty - I thought about it as a signal that the data is live (it is!).
I think it's a perfect (and subtle) way to signal that refreshing is unnecessary to see the latest data without wasting UI space explicitly explaining it. It was my favorite thing about the UI and I will be borrowing it next time I design a real-time interface where the numbers matter more than the precise timing.
really? it left a bad aftertaste for me
Re: I'm absolutely right
#155OK, so I love this, because we all recognize it. It's not fully just a tic of language, though. Responses that start off with "You're right!" are alignment mechanisms. The LLM, with its single-token prediction approach, follows up with a suggestion that much more closely follows the user's desires, instead of latching onto it's own previous approach. The other tic I love is "Actually, that's not right." That happens…
Yeah, I figure this is also why it often says “Ah, I found the problem! Let me check the …”. It hasn’t found the problem, but it’s more likely to continue with the solution if you jam that string in there.
We don’t know how Claude code is internally implemented. I would not be surprised at all if they literally inject that string as an alternative context and then go with the higher probability output, or if RLHF was structured in that way and so it always generates the same text.
Re: I'm absolutely right
#156Earlier quoted context omitted.
This can still happen even with thinking models as long as the model outputs tokens in a sequence. Only way to fix would be to allow it to restart its response or switch to diffusion.
I think this poster is suggesting that, rather than "thinking" (messages emitted for oneself as audience) as a discrete step taken before "responding", the model should be trained to, during the response, tag certain sections with tokens indicating that the following token-stream until the matching tag is meant to be visibility-hidden from the client. Less "independent work before coming to the meeting", more "mumbli…
Doesn’t need training. Just don’t show it. Can be implemented client side.
Re: I'm absolutely right
#157Earlier quoted context omitted.
I always find the claim hilarious that in relationships women are the ones who need to be appeased, when in reality it's mostly men who can't stand being wrong or corrected.
Gay male marriages remarkably have the lower divorce rates than heterosexual marriages, while lesbian female marriages have higher divorce rates. Multiple studies show that lesbians consistently have far higher divorce rates than gays. This implies a level of neuroticism with females, that they probably do need to be appeased more, and that if you have two needy people who need to be appeased it's probably not going…
Interesting. I know that gay men also make considerably more money than gay women, and having more wealth is associated with a lower rate of divorce, which sounds like a plausible explanation to me. I don't know if the numbers check out, though.
Re: I'm absolutely right
#158Earlier quoted context omitted.
I wouldn't go so far as to call this specific implementation a dark pattern, but it is misleading. It suggests the data updated right when I loaded the page, which obviously isn't true as I can see the same 16->17 transition on a refresh. I'd prefer a "Data last updated at " indicator somewhere. Now I know it's live data and I know how old the data is. Is it as cute / friendly / fun? Probably not. But it's definitely…
On that note, the font isn't symmetrical and the bar graph itself uses jagged lines. This makes it hard to read and much less precise. I'd prefer all websites in monospaced fonts with only the straightest of lines.
Those are stylistic choices that don't really impact the ability to view the data and do not mislead like the fake data update on page load.
You're able to hover a bar to see its exact value. Very precise there. No misleading info.
Re: I'm absolutely right
#159OK, so I love this, because we all recognize it. It's not fully just a tic of language, though. Responses that start off with "You're right!" are alignment mechanisms. The LLM, with its single-token prediction approach, follows up with a suggestion that much more closely follows the user's desires, instead of latching onto it's own previous approach. The other tic I love is "Actually, that's not right." That happens…
> The LLM, with its single-token prediction approach, follows up with a suggestion that much more closely follows the user's desires, instead of latching onto it's own previous approach. Maybe? How would we test that one way or the other? If there’s one thing I’ve learned in the last few years, it’s that reasoning from “well LLMs are based on next-token prediction, therefore ” is a trap. The relationship between the…
The human stochastic parrots (GP, not you) spouting these 2023 talking points really need to update their weights. I’m guessing this way of thinking has a stickiness because thinking of an LLM as “just a fancy markov chain” makes them feel less threatening to some people (we’re past the point where it could be good faith reasoning).
Like, I hear people say things like that (or that coding agents can only do web development, or that they can only write code from their training data), and then I look at Claude Code on my computer, currently debugging embedded code on a peripheral while also troubleshooting the app it’s connected to, and I’m struck by how clearly out of touch with reality a lot of the LLM cope is.
People need to stop obsessing over “the out of control hype” and reckon with the thing that’s sitting in front of them.
Re: I'm absolutely right
#160OK, so I love this, because we all recognize it. It's not fully just a tic of language, though. Responses that start off with "You're right!" are alignment mechanisms. The LLM, with its single-token prediction approach, follows up with a suggestion that much more closely follows the user's desires, instead of latching onto it's own previous approach. The other tic I love is "Actually, that's not right." That happens…
Is there a term when everyone sees a phrase like this and understands what it means without coordinating beforehand?
I would call it a meme