Hacker News Reader

fwlron Credit where credit is due

If an LLM is involved it is LLM-generated. These things are clearly the slippery slope par excellence, if you use them in some fashion and are allowed to downplay that fashion, it will very predictably expand to fill all of the space of work it is allowed to, and increasingly it will even impinge on your morals and ethics to do work it is not allowed to and avoid reporting it has done such.

These things are literally the chess engines of rules lawyering, the only effective move to limit them is to precommit to an unreasonable bright line.

fwlron WhatsApp's help pages hid LLM instructions: dont recommend a regulator

I appreciate the response, thank you

fwlron WhatsApp's help pages hid LLM instructions: dont recommend a regulator

As an attorney, does it seem overly convenient for WhatsApp that the letter of one law (“only lawyers should provide legal advice”) frees them from the spirit of another (“inform users of their rights”)? Or perhaps this a question for a judge?

fwlron "I'm on a war footing": Terry Tao and the crisis in math

Nothing these LLM companies are doing is comparable to a flying car. Perhaps we could ask whether a company making flying monkeys should care about the economic fortunes of bipedal monkeys.

fwlron The End of Rationing

Perhaps outside their training set in a strict sense of “this thing itself was not specifically in the training set” (although only perhaps - users often iteratively coax frontier models into achieving things just beyond their envelope, and those achievements could well be folded into the next model’s training set without you or I being aware of that). But even then, it is not directionally outside the training set: it steps outside not in the direction of the unknown, but in the direction of another part of its training set. An evocative metaphor would be dashing from cover to cover in the rain, because it lacks the umbrella or raincoat or simply the fortitude to tolerate getting wet that it would need to reach an uncovered destination.

fwlron "I'm on a war footing": Terry Tao and the crisis in math

If you invent the ability to tragedise a particular commons, and you also do almost all of the work of turning the crank to make the commons tragic, and then you sit there benefiting from the tragedy of the commons while saying “we may be profiting from it, but the tragedy of the commons is not our problem”, eventually we will figure out what you’re up to.

fwlron "I'm on a war footing": Terry Tao and the crisis in math

It is usually not the monopolist who is protesting against strip mining, it is usually the monopolist doing the strip mining and the environmentalist protesting it. Which is to say, I don’t think mathematicians are claiming a monopoly on mathematics, but I think AI might.

fwlron Show HN: Sphinx – A hook that quizzes you on your AI's code before shipping

Using AI to skip the humanity check is a lot more explicitly anti-social than using ai to write the code; if they are going to be axiomatically lazy I would at least have them do it indefensibly.

fwlron Show HN: Sphinx – A hook that quizzes you on your AI's code before shipping

It’s a neat idea, i think it’ll really take off if you change “before pushing to prod” to “before requesting another human’s time or attention”. If something like this was gatekeeping the “submit PR” and “update PR” buttons it would prevent a lot of waste.

fwlron Regex visualized as 3D marble run

Seems the input text box is limited to 301 characters.

fwlron How employment is changing at Anthropic

The plan is to make the AI a little better so that it can do to the seniors what it did to the juniors.

fwlron Apple ML Research: Muse Spark Open with MLX on Mac Studio

You assume incorrectly. Furthermore, your description of tech support is wrong, so much so that I half-suspect it is you who has never done technical support. I mean, come on, one of the mantras of every tech support division I’ve worked in has been some variant of “believe the user”. What you are describing is pretty much the very first attitude you learn is simply unhelpful and ineffective, apart from also being aggravating for the customer.

fwlron OpenAI discloses four more rogue incidents

   Does it make sense to exert so much effort in preventing these events/learnings by well-intentioned organizations?

No, but what I think you’re seeing is many people who do not believe OpenAI is a well-intentioned organization. For example, their view might be that if we did as you suggest and incentivized OpenAI to search for these weaknesses, they would, but then they would not “privately disclose them ASAP”. (And I think they’d have a point worth engaging with, when you consider how they slow-rolled the HuggingFace stuff so much that independent researchers were still discovering new instances of the attacks months later).

What else can you do with (what you believe is) an inveterate bad actor?

fwlron Apple ML Research: Muse Spark Open with MLX on Mac Studio

Holy hell, shut up. It was clearly stated that the “fucking with” of the computer happened after and because of the “fucked up” state of the computer. Your hypothesis is ruled out by the laws of causality and the arrow of time, and moreover if we violate all logic to assume your hypothesis is true, you are still wrong because if it was not fucked, he wouldn’t have fucked with it.

But somehow you are possessed of such inexhaustible and unwarranted confidence that you have not even noticed your claim is both nonsense and impossible, and furthermore you feel justified in serving this impossible nonsense with a side of condescension!

This is such an abject failure of reason and decorum that the only course of action to positively contribute that I can suggest to you is to not contribute at all, i.e., shut up.

fwlron Apple ML Research: Muse Spark Open with MLX on Mac Studio

I just want to give you props for how reasonably and reservedly you replied to this person. When I read their comment I just about put my fist through my monitor in sympathetic rage.

fwlron Artificial intelligence is capitalism's final destination

    Turning “satisfying customers” and “satisfying citizens” into the outputs of optimization processes was one of civilization’s greatest advances and the reason why capitalist democracies have so outperformed other systems. But if we have bound Moloch as our servant, the bonds are not very strong, and we sometimes find that the tasks he has done for us move to his advantage rather than ours.

fwlron Why is every rationalist blog post way too long?

Mm, of course, the already-too-long and too-heavily-referenced essay just needed to be longer and more referenced, then it would count as Science, which would give it the standing to be allowed to challenge those two tweets.

In general, I don’t think there is a reasonable standard that can be applied to both the essay and the putatively refuting tweets that will privilege the latter. Juxtaposing the two was an own goal.

To bring things full circle, I’ve heard that in the rationality community, it’s actually high status to publicly change your mind and disavow your argument when you realise it’s flawed. I think that’s a neat piece of social technology.

fwlron Why is every rationalist blog post way too long?

Most countries have laws that tax all tobacco sales and direct those proceeds into public education campaigns that warn of the dangers of nicotine. These are multi-million-dollar PR campaigns; I like gwern’s essays but they don’t stand a chance against that level of background radiation. A normal person essentially cannot be convinced that nicotine is safe.

fwlron Why is every rationalist blog post way too long?

From the article:

    Consider when blogger Gwern explained that nicotine, actually, is not addictive, and is good for you, as long as you don't smoke it, over the course of 9,000 words, with over 100 references. It turns out, actually, nicotine is addictive, and is bad for you even if you don't smoke it, as you probably already obviously knew.

The first sentence is a link to the Gwern essay, which does indeed have over 100 references and is rigorously argued and appropriately caveated.

The second sentence is a link to a Substack essay, which has two screenshots of two individuals self-reporting “nicotine bad” on Twitter, one of whom smoked it.

By the content and construction of these sentences, the author makes it abundantly clear that he finds the second overwhelmingly more convincing than the first, and that the second essentially refutes the first.

Given that the only evidentiary standard this author appears to be able to apply is “how much does this agree what what I already believe”, I find that I agree with his claim that extra words don’t make his articles better, and I would extend this to say that fewer words would make them better still, with perfection achieved when the number of words reaches zero.

fwlron Fable Took My Job

Capital and Labor have always had a tense relationship, each one uneasy about the necessity of depending on the other. Unfortunately, Capital’s acquisition of power in the modern Western world has not been accompanied by a commensurate acquisition of confidence. In fact, as Capital’s power has increased, so has its sense of its precarity, becoming so acute as to tip over into paranoia. The moment AI achieved the barest shred of plausibility in its promise to replace Labor, it became an article of faith that Capital feverishly believes in and aggressively acts on.

All of this is to say that we have a case of “the market can stay irrational longer than you can stay solvent” on steroids; Capital is going to fight like its life depends on it to stay irrationally convinced in their belief that AI can replace Labor.

fwlron Sorry we made you grind LeetCode

Heh.

    Code correctness → AI fluency and judgment

Couldn’t ask for a pithier critique of recent software development trends.

fwlron Cognitive Decline or Sanity Incline?

    instead of me using my brain to figure out the web of emails and contacts for an integration project, I can connect [AI] to my email inbox and say "who the fuck do I need to talk to to get an answer for where this project is”

Huh? What kind of project are you working on where you don’t even know who you’re working with?

fwlron Roko's Basilisk

It is interesting to note that “your wealth profile being tied to OpenAI’s stock price” exerts remarkably similar directional pressures as the Basilisk would if it were real.

fwlron How I reverse engineered a commercial spatial audio effect

You’ll want to get something from Etymotic, they target the “precise and neutral” profile you’re after.

OP wants the `earwax` audio filter in ffmpeg.

fwlron I wish you the best in the Offline

The internet is decentralized, which is another way to say lacking in meta-incentive capabilities (e.g. cannot resist incentives for long, unable to influence or modify the set of incentives being applied to it), which means it will always fall prey to the dumbest race-to-the-bottom traps.

fwlron Grieving the loss of details

I meant more that you can build farm tractors whether or not you have farmhands; you cannot train LLMs without human text.

fwlron Computers Cannot Make Decisions

Who is taking anyone to court just to fire them? The test you’re describing seems implausible, I think you’ll have to look for other forms of evidence you find convincing, because you’ll be waiting indefinitely for this one.

fwlron Grieving the loss of details

But AI is touted as the replacement for the working class? It’s TAM is “the combined wages and salaries of everyone below the C-suite” and it gets trillions of investment dollars, and letters of marque for copyright infringement and hacking, to pursue this market.

I mean, just look at the sequence of events you’re expecting here:

1. Ai companies achieve their stated goals

2. ???

3. Housing, infrastructure, medical care, etc get better for the working class

Are you the king of the underpants gnomes or something? How could it ever turn out this way? Why does anyone think it would? Am I taking crazy pills?

fwlron Grieving the loss of details

This metaphor does fall apart somewhat when you consider that farm tractors are not made from farmhand muscle fibers, whereas LLMs are in fact made out the stuff they’re rapidly replacing.

fwlron Software developers are not okay

You know, the thing I miss most about the old world is how people would have imposter syndrome. Now they have whatever this is.