tempfile

> It's not a person

and yet we as a species are spending trillions of dollars in order to trick people that it is very very close to a person. What do you think they're going to do?

That's true, but luddism is popularly associated with opposing useful technology (and is badly understood by most people anyway).

They are replying to someone who said ChatGPT. Why are you barging in to change the goalposts?

No. Graders having multiple roles is actually the implementation detail, since they're people, and they can't spend all day grading work. Scanning machines don't really grade work either, but I am happy to rely on them for checking an answer matches a scheme verbatim. I'm not sure why you mention scanners answering tests either, since my original comment doesn't imply that.

There is no evidence that an LLM can reliably evaluate the semantic content of a sentence, even in cases where we all agree that the semantic content exists. The thread we are participating in demonstrates a particularly egregious failure, but there is no good reason to think that more subtle failures might not exist if we happen to patch this one. Even if they were reliable, you can't evaluate a system with itself - that is basic science.

Not "control", "a control". As in a control group, for a study.

> Discriminating good answers is easier than generating them.

This is actually very wrong. Consider for instance the fact that people who grade your tests in school are typically more talented, capable, trained than the people taking the test. This is true even when an answer key exists.

> Also, human labels are good but have problems of their own,

Granted, but...

> it isn’t like by using a “different intelligence architecture” we elide all the possible errors

nobody is claiming this. We elide the specific, obvious problem that using a system to test itself gives you no reliable information. You need a control.

Isn't this one of the basic ideas of Lisp?

A machine that emulates a bird is indeed a mechanical bird. We can say what emulating a bird is because we know, at least for the purpose of flying, what a bird is and how it works. We (me, you, everyone else) have no idea how thinking works. We do not know what consciousness is and how it operates. We may never know. It is deranged gibberish to look at an LLM and say "well, it does some things I can do some of the time, so I suppose it's a digital mind!". You have to understand the thing before you can say you're emulating it.

So you think a rodent is a function?

The "point" of not anthropomorphizing is to refrain from judgement until a more solid abstraction appears. The problem with explaining LLMs in terms of human behaviour is that, while we don't clearly understand what the LLM is doing, we understand human cognition even less! There is literally no predictive power in the abstraction "The LLM is thinking like I am thinking". It gives you no mechanism to evaluate what tasks the LLM "should" be able to do.

Seriously, try it. Why don't LLMs get frustrated with you if you ask them the same question repeatedly? A human would. Why are LLMs so happy to give contradictory answers, as long as you are very careful not to highlight the contradictory facts? Why do earlier models behave worse on reasoning tasks than later ones? These are features nobody, anywhere understands. So why make the (imo phenomenally large) leap to "well, it's clearly just a brain"?

It is like someone inventing the aeroplane and someone looks at it and says "oh, it's flying, I guess it's a bird". It's not a bird!

I don't think this is right. The WCAG allows for "programmatically determined link context" which includes text surrounding the actual link. "click here" is bad but "Amaya website" or "Amaya" are fine.

e.g. from the WCAG examples you link to:

> An account registration page requires successful completion of a [Turing test](https://www.w3.org/TR/turingtest/) before the registration form can be accessed.

If how you manage your time is your business, why are you replying to an article about changing the length of the work week? You are telling us that nobody mandates your work week. Most employees are not in that situation.

> I can finish it on Monday and have it ready for the Friday deadline and then spend my time how I wish OR I can report the task completed and then expect to have more tasks assigned

Are you serious? Do you think you just have free choice of whether to do more work or not? If you completed a task in 15 minutes and spent the rest of the week surfing do you think that would be allowed by your employer? Most employee contracts would forbid it.

> It is none of my employer’s business how I spend that time as long as deadlines are being met

This contradicts what you already said ("there are always more TODOs"). Which one should I reply to?

If your employment has fixed deadlines, and your employer does not react to efficiency increases by setting earlier deadlines, then you should expect the headline effect (efficiency increases mean your weekly hours decrease).

Not exactly true. The typical work week was longer in the 1800s than in the 1900s. The limiting of the work week to 40/5 was an explicit political movement. Of course it was not reduced as a direct consequence of increased productivity, which I guess is what you meant.

Do you come in on Saturdays, since your TODO list is still not empty?

Because the enterprise is not owned or controlled by you, and you don't get to set the rules. If you can negotiate better conditions, and nobody else on the market can undercut you, then you can have them. This isn't even an abstract consequence of capitalism, it is an axiomatic, defining feature. If you don't like it, you are rejecting capitalism.

and that would be a bad thing?

XSLT is probably the #1 reason people get turned off from XML and swear it off as a mistaken technology. I actually quite like XML, so I have been trying lately to tease out exactly what it is that makes XSLT a mistake.

XML is a semi-structured format, which (apart from & < >) includes plain text as a more or less degenerate case. I don't think we have any other realistic format for marking up plain text with arbitrary semantics. You can have, for example, a recipe format with <ingredient> as part of its schema, and it's trivial to write an Xpath to pull out all the <ingredient>s (to put them in your shopping list, or whatever).

Obviously, XSLT is code. Nobody denies this really. One thing about code is that it's inherently structured. Only the craziest of literate programmers would try to embed executable code inside of text. But I don't think that's the biggest problem. Code is special in that special purpose programming languages always leak outside the domain they're designed for. If you try and write a little language that's really well-scoped to transforming XML, you are definitely going to want to call stuff outside it sooner or later.

Combined with the fact that there really isn't any value in ever parsing or processing a stylesheet, it seems like it was doomed never to pan out.

Still rocking a 3 here (albeit only 5 years old). Going quite strong after a couple of replacements and upgrades.

But you compared it to vim in the project description. If you don't think it's better than vim in any respect, why did you write it? I am only looking for reasons I should try to use it. You seem to imply there are none?

Spent 15 seconds waiting for the spinner to resolve before I gave up.

Looks lovely. Where does it deviate from vim? Evidently it is modal. What features make it more effective than vim is?

> You make it sound like systemd manufactured this monoculture somehow.

Where are you getting this from? I do not see it at all. The parent comment just says that it is an emergent compromise they don't think is a good one. That the code cannot be audited is also not necessarily a quality issue, either. It is just impossible to feasibly audit over 10 million lines of C. (this criticism applies equally to the kernel, although I doubt anyone would claim the kernel is less audited than systemd)

> Let's not pretend that a modern "init" does only system initialisation by calling shell scripts and then disappearing.

Nobody is pretending this. the comment you are replying to literally says "I don't think it represents the current state of the community".

What does that actually mean? The tweet is not very clear. Is it the bold in the header? What's special about it?

I have not been able to configure an editor that makes it less painful. I think it's not just the verbosity (although that is painful) - the prefix notation makes some things inherently hard to read. MathML is almost impossible.

I have been thinking about this seriously myself. Not with a specific existing schema like DocBook, but with a custom schema (defined by me) that I then compile to standard schemas, like DocBook or HTML.

This seems extensible to the degree that I want (i.e. semantically rich enough that you can conceivably hang any application from it). But I just can't bring myself to write in XML syntax, especially for maths.

Jesus, who writes this stuff?

> AI is unavoidable

> We will speak to models in voice mode, chat with companions for 2 hours a day, and point our phone cameras at our surroundings in video mode

This is surely meant to be an objective assessment, not a fluff piece.

That is helpful to me! How did it work? I imagine a lot of RC interactions are kind of spontaneous, and I never understood how that would translate over the internet.

I'm not going to the US - that is the entire point of my comment.

Is asking about Recurse Center on-topic? I wanted to apply, but I genuinely don't consider it safe to travel to the US. As far as I can tell there is no comparable bootcamp-style retreat in EU :(