@4bpp's banner p

4bpp

Now I am become a Helpful, Honest and Harmless Assistant, the destroyer of jobs

3 followers   follows 2 users  
joined 2022 September 05 01:50:31 UTC

<3


				

User ID: 355

4bpp

Now I am become a Helpful, Honest and Harmless Assistant, the destroyer of jobs

3 followers   follows 2 users   joined 2022 September 05 01:50:31 UTC

					

<3


					

User ID: 355

The most recent example I can think of is Zed, a "work with LLM agents" IDE written in Rust that has been getting a lot of traction (and seems to be the only way around to do this with a GUI if you don't want to touch Electron). I gave it about two days' college try since I'm on the market for a new IDE (after CodeLite got enshittified with a recent update). Specific examples of suck about it:

  • Quiet or inexplicable failures. If it doesn't like the way the LLM it connects to formats something, it either leaves you with an unactionable error message ("missing field 'path' in tool call") or outright does nothing (such as when it doesn't like the structured text format it gets back from the LLM it connects to for its "smart autocomplete" feature - a state that is indistinguishable to the user from "I failed to connect", "the suggested completion is empty" and "the model is still working"). To debug any of these things, you have to get the raw JSON data it received out of an undocumented sqlite file, and guess what the expected schema is, because that is also undocumented. (It's hard to figure out even from looking at the source, because the JSON data seems to be fed into the handling Rust class by some weird metaprogramming magic rather than an explicitly defined format.)

  • Sometimes parts of it, like the LSP integration, just stop working until you restart it. Again, no visual indication of failure state, the feature just doesn't fire anymore.

  • It tries to do something clever with pending edits, but there is no coherent underlying model. A lot of LLMs clearly assume that they can keep drafting and re-drafting an "edit file" tool call, like submit one that says "add a line saying X between the line that says A and the line that says B" and then follow up with "add a line saying X and a line saying Y between A and B". The latter fails because for the purpose of that tool the state is now A\nX\nB, but if the model tries to read the file again to get a consistent picture it still only reads A\nB.

  • Clearly there is lots of internal code duplication, resulting in non-orthogonal config duplication. There are separate configurations for the "agentic" LLM connection and the "autocomplete" LLM connection, supporting a slightly different subset of protocols, with slightly different design choices e.g. about whether you give the top-level URL of the API endpoint or the most specific one.

  • There is no build system integration whatsoever (even though this is the sine qua non that distinguishes an IDE from a fancy text editor), but the lede is buried about this. I get the sense that they want to have this eventually, but are just not too excited to work on it, while not wanting to pay the marketing cost of outright saying "we are not ready for prime time yet".

  • There is debugger integration, which is supposed to be able to do fancy modern things like letting you see variable values by hovering the variable in the source as you debug. This works about 10% of the time, again with no indication when it fails.

All of this hints at a level of engineering that, if it were C/C++, I'm sure would result in a program that is unusable due to segfaulting all the time, with a reputation to match. In Rust, it instead just keeps running, slowly piling the frustration on you, while the developers psychotically add one shiny new feature after the other and leave the occasional bug reports about the above to rot without a response (because who cares about one user's unverifiable report that some asynchronous thing didn't appear? It's not like they can bring receipts in the form of a stack trace, because there was no crash).

Earlier, the Ladybird web browser was also rewritten in Rust mostly by LLM.

Oh. RIP. "You were supposed to be the chosen one", or something.

Maybe it's my eternal grumpy old "real programmers should write in..." fart eyes biasing me, but every single major project written in Rust that I have tried so far turned out to suffer from the same problem as major projects written in Python and JavaScript did before: it doesn't crash, but it sucks in more subtle ways that make me think that (1) if the language didn't prevent that, it would probably crash every 5 minutes and (2) if it did that, its developers and defenders would at least realise how bad their program is. It's strange because when the "it's memory safe and allows you to focus on the important things!!!" language was Python/JS, the obvious theory to explain this was that the language enables incompetent programmers who are not actually up to the task of engineering big pieces of software; but Rust is not even particularly easy to write, and arguably harder than C++ (unless you fully commit to the early-2000s high corporate obnoxious_template_factory<namespace::bullshit_factory<allocator_factory<...>,...>> style). Still, the filter fails. Maybe the missing puzzle piece is just the AI: it has no trouble emitting walls of lifetime sludge that passes the compiler, but is no better at big-picture decisions than the Python script kiddie, all the while its script-kiddie-in-mindset handler looks at it and figures that because it is Rust and it runs, the program generated is definitionally beyond reproach.

I think it's a similar question to "why do people who clearly know the difference still don fake Prada/Gucci/Rolex apparel". To a lot of people, not being able to "write good" is a lifelong source of feelings of inferiority. AI posting both offers the promise that they can finally get some of the respect that the verbally gifted command (whether justifiedly or not) for themselves, and a bit of a sense that they are getting their cheeky revenge on the system that was keeping them down.

2 is basically saying "if you do this you will lose the culture war". Isn't the whole point of the "don't fight the culture war" rule that you are not allowed to act like you are primarily motivated by winning it? So maybe this is a strange place to be posting if you are swayed by that point.

I don't know if surrogacy would be particularly more popular on the Left in Germany; in fact if I had to make a blind guess I would imagine the FDP to be the party whose electorate is most okay with it, since it really registers as more of a "affluent people defying nature and propriety" vibe than anything with obvious left/right implications. If the CDU makes it a right-wing thing, they are just rationalising their impulses using whatever their chosen branding is.

I'm about as opposed to the Israelis' actions in Palestine as it gets, but I believe that hosting the UN ought to entail some sort of responsibility that surpasses object-level moral concerns. Really, the UN is somewhat like the Motte in that its purpose is to provide a neutral ground for parties to talk to each other who would never do so otherwise, except with the UN there are rather more things at stake.

If the people of NYC can't keep their moral principles in their pants for this higher purpose, maybe they should in fact not be allowed to host the UN, and the headquarters should move to a place with a higher decoupling waterline. It's their loss, or that of the US as a whole if they are incapable of finding any other site to guarantee safe passage for the de facto representatives of every significant piece of clay. (I'm sure China would be happy to set up a special economic zone with hookers and blow somewhere if nobody else is willing to step up.)

So you figure every disgust reaction that exists in humans [in significant numbers] is natural for all of them? In that case, it seems like you are positing that the state of nature is to be disgusted of quite a lot of things, perhaps enough that human functioning is very nearly impossible without training yourself out of some of them. If nature requires that you train away some of your disgust reactions and society determines which ones, is it really fair to say that the ones you happen to keep are all natural and not socially conditioned? I don't know if that helps the case for "gays bad because it is natural to find them disgusting", either.

Not to mention I don't think you can socially condition people into finding things gross

I can't, society at large surely can. Apart from the examples that already came up, there are the food cases of raw fish, animal meat, insects, milk and a variety of fermented products, of which I'd wager the majority of humans on the planet experiences disgust reactions to at least one of. How does your model of disgust reactions as something natural explain that?

Well, a wordcel beloved of many here once quipped that the sovereign determines the null hypothesis, and the current sovereign surely is not on the side of aversion to homosexuality being natural or inevitable.

More prosaically, if what you are seeking to prove is that homosexuality is unnatural, then the outcome that gut responses are not against it due to nature is a null result.

Yet on the other hand people especially on the US right still regularly invoke the memory of the 1979 Iranian hostage taking of US diplomats. How important an event seems to remember greatly depends on how well it meshes with the narrative you are weaving.

(Also, isn't the consensus nowadays that the Gulf of Tonkin Incident was a US intelligence fabrication, almost up there with Sender Gleiwitz?)

I feel like this came up in the previous discussion too, but there is the almost uncontrollable null hypothesis that it's still all subtle social conditioning (since it seems that disgust responses towards interracial relations were once similarly common but now I wouldn't be surprised if even a bona fide KKK member would struggle to muster visceral disgust at a WM*F couple making out). Of course if true this is a boon to the progressive case for deep social engineering.

I imagine that in the eyes of many in those organisations, this is just some form of "only the first violation is bad" ethics. (Absent a functioning state to prosecute for murder) it is okay to murder a murderer, and likewise (absent state punishment of stochastic murderers) it is okay to stochastically murder a stochastic murderer.

I guess I'm mostly talking about European airlines here. I guess the US generally has more of a "money is cheap" (when compared to the loyalty of a high-status customer?) attitude.

Eh. The fundamental argument that if you respond to "Country X occupies the land and immiserates the people" by taking in all the people and giving them a better life, you make country X's job very easy and hasten the opportunity for them to do it to you next, is sound enough, both for X=Israel and X=Russia. The fundamental difference here is that the Eurocrat lobby assert their mandate over anyone who would stand against them, within or without their dominion, by claiming moral superiority as the defenders of universal individual human rights, and to my best knowledge Egypt does no such thing. By doing this, but then still turning around and making their military convenience trump those same notions they just declared universal (e.g. when browbeating the internal opposition that takes issue with their importing of Muslim refugees), they are guilty of hypocrisy. In my eyes, official hypocrisy, being a meta-level crime against the very concept of principles, is a worse crime than any object-level violation of principles can be.

(Those who build moral frameworks to protect hypocrisy, like the concept of "whataboutism", get close to being yet another level of abstraction up, thus deserving their own deeper circle of hell.)

The vast majority of my customer service interactions has been with various transportation companies (airlines most frequently), who very much do appear to optimise for dodging refunds that their customers are in fact legally entitled to. There is a reason "pay us a third of the refund and we will take on the effort and risk of enforcing your statutory compensation claim against the airline" is a real business category that exists out there.

The problem with customer service (and a lot of other similar domains, actually) is that as the problem to be dealt with rather than the employer, the human bottleneck actually often worked in your favour. Having human employees working the phone line and wanting them to not quit or flame out and shoot up the office is the fundamental limit that makes it hard for Corporate to institute their ideal customer service policy, which is "trap any complainants in a Kafkaesque gaslighting nightmare until they give up". Mr. Claude has no limitations there, because he does not feel the "I am screwing over a fellow human being and making a mockery of the very concept of 'support'" qualia nor the "it sucks to be screamed at all day by people who hate me" ones.

I don't know. Many societies (e.g. Germany) have decided to not acknowledge a right to contractually sign away your bodily integrity (e.g. you can't sell your arm, while you are alive; the person taking it will be criminally liable for the mutilation regardless of your consent). I don't think there are obvious reasons why this is bad, beyond an easy-to-draw Schelling fence around "consenting adults shall not be restricted in their dealings with each other" which never actually existed anywhere. Once you have this, though, making reproductive autonomy another thing that a woman may not contractually sell is not a far leap.

...but doesn't that mean you don't in fact make the distinction? You seem to be saying that the cases of undesirable outcomes I mentioned (guns, procedural acquittals) are in fact all instances of slippery slopes, and there is just nobody making the claim that expecting the slope to be slippery is fallacious.

(When I say "slippery slopes", I really do mean what I understand that lemma to mean, not shorthand for "instances of the slippery slope fallacy". My view is that neither of the three cases in discussion (abortion, guns, procedural acquittals) should count as an example of having slipped down a slope, because there was no widespread slipping - guns caused a rounding-error number of mass shootings and still approximately nobody likes mass shootings, and legal abortion plus surrogacy are now producing a rounding-error number of surrogates being pressured to abort, which also approximately nobody likes)

I'm happy to accept that refusing to publicly concede that there is such a thing as a tradeoff at all, and their policies might not be all upsides, is an endemic vice of US progressives as well as many other movements that code left. I'm not sure if it's so much a left-wing thing as it's a "progressive and in power" thing, where if you have some utopian vision that you are already ramming through and just want to ram through faster against the feeble protestations of a browbeaten conservative opposition, conceding that there might be downsides at all is tantamount to surrendering to those voices who are begging you to slow down and reconsider (and you suspect are just angling for some more time to extract value under the curve of their evil ways before they are reduced to zero).

Regardless, do I understand correctly that are you saying that this is really a "slippery slope" rather than an "unfortunate consequence" specifically because the hypothetical progressives beforehand said "this won't happen"? (and if they said "it will happen but that's a good thing"/"worth it" that would be enough to make it an "unfortunate consequence"?) Not only am I not sure I've seen evidence that the proponents actually addressed this specific case, it seems a bit strange to me to pin such a great distinction in evaluating a policy and its consequences not on anything about the policy and consequences, but instead on the honesty of the propaganda of its proponents.

It absolutely is because it never could have happened in the first place without the prerequisite social and legal changes at the top of the slope.

This is like the entire point of the slippery slope as an argument.

Are you willing at all to make a distinction between "slippery slopes" and "unfortunate consequences"? With the particular case at hand, I doubt a lot of progressives are particularly happy with what is going on either; it's just that they would pin their displeasure on the surrogacy aspect of it, not the abortion one. If almost nobody wanted an outcome and it happened in one instance, is that enough to make a slippery slope? In that case, there are really a lot of slippery slopes everywhere. Is every gun murder and mass shooting is at the bottom of the "slippery slope" that started with the 2nd Amendment? Is the entirety of the Anglo-American legal system a slippery slope that led to OJ Simpson walking free? If not, what's different there?

the past

Actually, I have to apologise there - I expected to find a particular shape of scenario (like "some nobleman forces his mistress to get an abortion") but after putting in some time to search it did in fact not seem to have occurred (at least not in a way that left any evidence known to us today). The closest-in-vibes stories I can find are Henry VIII/Anne Boleyn, a variety of "queen used her legal authority to torture pregnant mistresses" stories, and the whole lot of sexual violence in American slavery (though I guess someone very upset at abortion in particular would find compelled impregnation/reproduction much less reprehensible than compelled abortion?). Mea culpa, my claim that similar things must have happened in the past was unsupported.

It's not evidence of a slippery slope if someone, somewhere, did one thing that maximally enrages you, out of hundreds of millions of opportunities for people to do such a thing. Do you see evidence of an epidemic of surrogates being pushed by their clients to abort? Do you have any particular reason to believe that things like that, or things that you would find worse than this, did not happen in the past? My vague impression of humanity in general and European history in particular (which I am of course more familiar with than the totality of humanity) is that there was never a shortage of any of the ingredients here, like sexual compulsion, inferiors being made to bear children for superiors, abortion and actual infanticide out of convenience. It seems very likely that something like this would have happened many times in the past, say, some high-status man impregnating a domestic worker or slave and then using his superior status to force her into abortion or to kill the child after birth. But if it did, when exactly did we supposedly slip down a slope?

This theory of ownership seems bizarre to me. If I convince the local paper to run an article saying my house is much nicer than my neighbour's, and this results in his property value going down while mine goes up, am I stealing from him?

The thing is that with such a loose definition of "in the training data", the hypothesis that AIs will only be able to do what's in the training data is not reassuring against doom. Persuasive propaganda is in the training data. Mass murder is in the training data. Deadly diseases are in the training data. World wars are in the training data. Doing all those things hundreds of times faster and cheaper than humans, like the current set of programming and science tasks where AI doing them faster and cheaper is being dismissed as uninteresting because it was all in the training data, would be more than enough to largely end humanity.

Is this like the common left-wing ideation that you could just take Elon Musk's wealth and solve world hunger, the housing crisis and grandma's medical bills all at once? A quick search says that USAID funding was on the order of tens of billions of $ in a year, while the Obamas' net worth is 70 million. You would need to reduce about 500 Obamas from caviar to dumpster diving to fill the gap for a year.

I do have the feeling that politicians tend to be able to draw on remarkable reserves of ability to compromise and act cooperatively when their personal interests are actually threatened (as they would be by randos with nukes). The wild defections you are talking about seem to be the province of things the electorate may care deeply about, but the politicians themselves are happy to game.