Hacker Newsnew | past | comments | ask | show | jobs | submit | drabbiticus's commentslogin

I really wish people would define terms when using math. What is y? What is LLM(x)? Presumably it evaluates to some real number so that it can be fed to the logistic sigmoid function. If it is the logistic function, then why does beta going to infinity matter? It seems to just collapse the output of the sigmoid function to 1 and make the value of LLM(x) meaningless instead of their claim that it recovers the LLM classifier. What is the function I()?

Maybe these are well understood terms in some field? Maybe I'm just lost?


x and y are common shorthand for inputs and outputs respectively. Saying "y=LLM(x)" denotes that the input tweets (x) are fed to a black-box function (LLM) that returns a classification (y) for that input. Given that we're talking about binary classifiation it's expected for y to be a value between 0 and 1, which leads to defining LLM(x) as "a function that receives a tweet x as input, uses an LLM, and returns a value in the [0, 1] range representing its classification as ironic or not".

The point about beta (which I think the equation doesn't actually reflect) is just to indicate that this approach risks nothing because, worst case scenario, the weights you assign to the model can simply return the original LLM prediction. Keep in mind that 0*inf=0 and that LLM(x) only returns 0 or 1, so it's not accurate to say that the sigmoid would always return 1.

The function I() is the indicator function [1] which, in this case, returns 1 if the argument is true and 0 otherwise. It's only there to convert booleans to integers because summing booleans is not defined.

[1] https://en.wikipedia.org/wiki/Indicator_function


Thanks, a lot of this helps.

Having said that, if

    - y is a binary "ironic"(1), "not ironic"(0) label and
    - x is the content of a tweet and
    - LLM(x) is meant to represent some probability range [0,1] representing p(y=1|x)
then it seems that p(y=1|x) === LLM(x) already.

Why bother with p(y=1|x) = σ( α + β*LLM(x) ) as in the article? You are right that I missed the case where LLM(x) = 0, but that just means that p(y=1|x) has exactly 2 values when defined as above and β -> inf.

    - when LLM(x) = 0 then p(y=1|x) = 1/(1+e^-α)
    - when LLM(x) = (0,1] then p(y=1|x) = 1/(1+e^-inf) = 1
This is substantially lossy and converts a continuous output LLM(x) of range [0,1] to a step function not even defined as a set {0,1} but instead the set {1/(1+e^-α),1} for unclear gain. It also loses all sensitivity to LLM(x) near 0, and only detects/changes value when LLM(x)=0 exactly. It might make more sense if LLM(x) is not limited to [0,1] like you have claimed because then at least the logistic regression is clamping the output to [0,1]. If you were wrong about the range of LLM(x), this would come back to asking authors to actually define their terms.

My point is that when math is used to justify something or communicate something, it should be explained or very apparently right. If when someone goes to try to understand the math it doesn't match the claims being made in the article ("recovering the LLM") then it throws the rest of the article into doubt.


I don't think that the Anisakis species discussed in this article would generally be considered microscopic, as demonstrated by various photos included in the article. The Trypanorhyncha larvae are smaller, but also likely to be visible without microscopy as well, on the basis of the scale I'm inferring from the included photos vs. the muscle striation in the tuna.


12 hours before this was posted to HN: https://grapheneos.social/@GrapheneOS/117198980596943121

GrapheneOS: "Multiple Google engineers we've contacted have said they aren't able to give us any information about this so we're left doing reverse engineering and relying on leaks. The leaks do not seem reliable and do not match what we see. Our concern is that MTE may actually be broken due to CPU errata."


I would argue that "works fine" != "direct support" when discussing vendors and platforms. Support is a commitment signal, not just a technical capability.


Seems likely to be another example of LLMs replicating training corpus. My expectation is that there will be more writing showing why something is counterfeit than why something is genuine, in part because disproving that something is counterfeit from visual/physical clues alone is essentially impossible. Lack of any definitive counterfeit proof is not proof of authenticity, hence why provenance and trust as such integral market factors. For example, even if the packaging were to be identical (e.g. say it's stolen from the factory), this is still not proof that the contained product is the one advertised.

And yes, packaging mistakes with chemical names and typography are quite common for anyone who occasionally pays attention to these things.


This seems likely to be cultural, possibly even microcultural.

I know many adults who believe that splitting bills is a normal and reasonable thing to do.


Splitting bills, where each pay a part in the restaurant, fine. Splitting bills where one pays and the other sends him money, not.


Who is making these rules up? Why do you think they apply universally?


God made them, as far as I know. Humanity lived in gift economies long before the invention of barter and trade.

These are etiquette rules, which have their benefits. Since nobody is forced to follow etiquette, they help you to learn things about people.

Generosity is one of the easiest life hacks to find out who is friendly towards you and who is not. If you buy them a drink and they won't buy you a drink back, that's a tiny cost for you to know what kind of fellow you're dealing with. If they buy you a drink and then asks you to transfer money to them, you also learn something about them. If people splitting bills want to sit and count to make sure that everybody pays exactly for what they had, then you learn something. And so on.

If you invite people for dinner a few times, you will learn something about that person whether you wanted to or not.


The presence of social constructs is advantageous to group survival, and an individual operating within a given set of such constructs may choose to adhere to or reject the norms of those constructs. As you say, you do learn things about people through interacting with them. I think ideally this interaction includes discovering what their norms are, whether they believe they are conforming/rejecting/otherwise to a status quo, and the reason for it. The action itself is may not be sufficient to infer an individual's values without the context and reasoning, especially in the absence of clear communication. Social constructs are local to social groups and subgroups, with differing levels of conservation and generalization between groups.


> quite a bit younger than

You have introduced a comparison, which thus begs the following questions:

1. What is the event that defines "China proper"? 2. What is the event that defines "Europe [proper]" to you?


1. Whatever was happening in Zhou. (Not sure it was really a "dynasty" pe se.)

2. Proto-Indo-European and Ancient Egyptian cultural norms and patterns are still very much alive in "Europe" today, so probably the invention of the wheel in the early Bronze Age is a good start.


I don't think Firefox should automatically spoof a Chrome user agent.


There's a long tradition of browsers doing exactly that.

The Chrome user agent starts with Mozilla/5.0, ends with Safari, claims to be like Gecko, based on KHTML, and somewhere in there is the string Chrome.

That was done exactly because people were trying to gate which browsers could or couldn't see their page by name, instead of by missing functionality.


If a site doesn't work, I disagree. It should.

The web is becoming a monopoly anyway... a contradiction to what web once meant.


Not even saying you are wrong, but the linked article does not appear to support your premise that AI was related to this decision at all.

Is there further evidence you have that specifically relates to the specific decision of this particular dev to step down from Arch?

As you yourself say, being a maintainer is a big commitment, one that many people naturally tire of as their interests, goals and other life needs change over time.


Phoronix has some more info, AI isn't mentioned: https://www.phoronix.com/news/Arch-Linux-Foxboron


What supplemental information is there ? I did not read much more than in the linked article.


The Phoronix article contains no more information than his resignation email.


Just imagine a "Missing Milk" [1] on the side of a milk carton campaign. I'd be so confused haha.

[1] https://en.wikipedia.org/wiki/Missing-children_milk_carton


Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: