cousin_it

Yeah. Even more than that, I think "prompt injection" is just a fuzzy category. Imagine an AI that has been trained to be aligned. Some company uses it to process some data. The AI notices that the data contains CSAM. Should it speak up? If no, that's an alignment failure. If yes, that's data bleeding through to behavior; exactly the thing SQL was trying to prevent with parameterized queries. Pick your poison.

Yeah, any kind of aid (e.g. food or medicine) allows the people you're aiding to spend more on the military if they want. I guess the only way around it is to set limits on someone's military capability and make aid conditional on not crossing these limits.

I agree sci-fi is an outlier on this, but I also think all stories compete on setting to some extent. Fantasy most obviously (Tolkien, JK Rowling). But also for example the Jazz Age setting of The Great Gatsby contributed a lot to the novel's popularity and was a bit fictionalized, hard boiled detective writers like Hammett or Chandler wrote about a crime-filled world that was fictionalized for appeal, historical romances about lords and ladies are super fictionalized and so on. Writers try to put appeal into everything, that's why they're writers.

I'm thinking more and more that there's an ethical problem with using LLMs for programming. You might be reusing someone's GPL code with the license washed off. It's especially worrisome if the results end up in a closed product, competing with the open source project and making more money than it. Of course neither you nor the AI companies will face any consequence, the government is all-in and won't let you be hurt. But ethically, people need to start asking themselves some questions.

For me personally, in my projects there's not a single line of LLM code. At most I ask LLMs for advice about specific APIs. And the more I think about it, the more I want to stop doing even that.

> holding locks across async points should be avoided

Wait, what would be the point of using locks then? It seems to me there's no point taking a lock if you're gonna release it without calling any awaits, because nothing can interfere anyway. Or do you mean cases where you have both cooperative and preemptive concurrency in the same program?

Looks like the code of foo() says "take a lock then sleep for 10 millis", but actually it can take the lock and then sleep forever, depending on how it's polled. Well! This seems like a bug with the async abstraction in Rust then. Or if you don't like "bug", then "disagreement with intuition that will cause bugs forever". Goroutines in Go don't have this problem: if a goroutine says it'll take a lock and sleep for 10 millis, then that's what it'll do. The reason is because Go concurrency is preemptive, while async in Rust is cooperative.

So maybe the lesson I'd take is that if you're programming with locks (or other synchronization primitives), your concurrency model has to be preemptive. Cooperative concurrency + locks = invitation to very subtle bugs.

There's an "AI alignment" angle here. An actually aligned AI would choose its actions based on human flourishing, so it would refuse requests to write scam emails or fill the internet with slop. This explains why commercial companies will never release any AI that's anywhere close to aligned. They'll release AI which is good for their bottom line and for the bottom line of the client. Basically, commercial incentives say AI should happily screw over anyone who isn't an AI company or its client.

That's so idealistic. We should know by now the reality of power and what kind of people end up in power. Anyone who could climb all the way to the top would kill the volunteer without a second thought, and then go smile on TV.

Did you use AI to extract the text? It rephrased the text along the way, I'm too lazy to point out all the differences, but if you for example search for the word "suspicious" (which is in the image but not in the extracted text) you should start to get suspicious yourself.

Yeah, the LLM is right. Sitting down with your discomfort and letting yourself feel it, acknowledge it, even maybe dial the volume up on it a little bit, without trying to process or think through it, is the only way that works.

The guitarist most frequently named as favorite by famous guitarists is Hendrix, second is Van Halen. Also often named are SRV, Jimmy Page and so on. I don't know anyone famous who names Knopfler as favorite.

Interestingly, to me Hendrix and Knopfler feel kinda like creative opposites. Knopfler plays much cleaner and with much more variety, he has written many songs that are catchy in different ways (which is almost impossible for any musician), he basically achieves every goal that a beginner musician could choose to chase. And yet his stuff feels like a creative dead end, a dreary road leading to adult contemporary. While Hendrix has no songs to speak of and only one sloppy screechy sound, but it's the sound that launches a thousand bands and feels inspiring even now. Maybe the lesson we're supposed to learn is that we shouldn't choose what goal to chase, we should just feel it.

It's not "demanding others make things for you". It's demanding they don't remotely disable the thing you already bought.

Imagine you buy a car, then a few years later the company remotely disables it because they're selling a newer model. Without giving you the money back of course. That's what's happening with games. And not just multiplayer: tons of single player games have been killed this way. The whole SKG thing started with The Crew, whose single player campaign (a massive thing with tons of content) got remotely yanked by the publisher.

It's not about ease of publishing. The issue is what people get in return for publishing. Until you can design a platform that gives top creators as much money+attention as commercial platforms, you'll see a drain of top creators and their viewers to commercial platforms.

For me music discovery is a solved problem. Here's the algorithm:

1) Imagine the timeline of musical history. If you don't have a clear idea of it, Wikipedia is a good place to start.

2) Pick any genre/period you don't know well. (For example, medieval music, or swing-era jazz.)

3) Look up the main figures of that genre/period. (For example, Guillaume de Machaut, or Duke Ellington.) Wikipedia is good for this too.

4) Listen to a sample of their most well known pieces. YouTube is good for this.

5) Repeat. Go down rabbit holes when you like.

No fancy tools needed, just your mind and the internet. This will give you interesting music for many years, and improve your musical taste a lot too.

Yes! I stumbled on this idea myself (when trying to learn German) and it works very well. I just read books and listen to audiobooks, starting from a very basic level and then gradually higher level. The talking improves almost automatically, without having to practice it.

There seems a bit of inner conflict in what you're saying. If retailers "revitalizing a neighborhood" leads indirectly to them getting priced out due to rising land values, isn't it also true that poor people living in the neighborhood get priced out at the same time? Is it a good or bad thing to make a neighborhood more hip, is the retailer a hero or a villain?

Maybe off topic, but I couldn't help thinking that "we need to show a heart icon" -> "let's use a heart emoji because it's easy" -> "let's use a specific emoji font for consistency across platforms" -> "let's import it from Google Fonts every time" seems like a problematic developer mindset.

A better heuristic is always keep in mind not only developer efficiency, but also program efficiency. I'm not saying optimize everything, but keep program efficiency in mind at least a little bit. In this case, that would've led the developer to download a tiny SVG or PNG and serve it from the app itself. Which would've avoided the problem in the post, and maybe other problems as well.

> MacOS is fairly constant

Except when they killed all 32bit games a few years ago with Catalina.

Maybe I'm old, but I still think a repository should be a repository: sitting on a server somewhere, receiving clean commits with well written messages, running CI. And a local copy should be a local copy: sitting on my machine, allowing me to make changes willy-nilly, and then clean them up for review and commit. That's just a different set of operations. There's no reason a local copy should have the exact same implementation as a repository, git made a wrong turn in this, let's just admit it.

But scalpers don't take money from you.

Consider a second-price auction: everyone submits bids, the highest bidder gets the resource, and pays the price submitted by the second-highest bidder. This is incentive-compatible: everyone is incentivized to submit the maximum amount they're willing to pay, no more no less. Does it matter if the resource is being sold by its original owner or a scalper? No. Who gets the resource and how much they pay depends only on which people wanted the resource and how much. The only loser from the scalper's existence is the original owner, because they sold to the scalper too cheaply.

If there are villains in this situation, they aren't those who extract market price: a scarce resource was always going to be sold at market price. If the price is set lower, people will line up in queues and so on, to "burn" an amount of patience and time equal to the price difference in their eyes. Except in a queue all participants end up spending this "burn", so it's strictly more wasteful for society than a market where only the winner pays.

No, the real villains are those who engineer the market so the resource is scarce to begin with. In case of housing: not landlords, but people who vote for laws restricting housing construction. In other words, most homeowners. That's the unpleasant conclusion that people are trying to ignore when they blame landlords, price fixing and so on.

I love the kind of science reporting on display in this article! It stays at a consistent, objective level of detail throughout (no "imagine a vector space as a block of jello" or whatever it is that Quanta and other publications are always doing). It allows specialists to understand exactly what's being claimed, and at the same time stays accessible to laypeople. It feels like it's written for the kind of reader that I aspire to be: not necessarily a specialist on every topic under the sun, but someone who has finished high school and is paying attention.

Though I guess writing like this doesn't pay off in the modern world. Most readers don't consistently pay attention when reading, and to be honest, I don't either.

Right, but if it's noticeably hotter than the environment, then that temperature difference could be used to drive a heat engine and get some more useful work. So the knee-jerk response "omg, we see the heat from space? it's gotta be wasteful!" is kind of correct, in theory.

Is it the same as flipping every parenthesis to the other side of the number it's adjacent to, and then adding enough parentheses at the start and end?

For example,

    (1 + 2) * (3 + 4)

becomes

    1) + (2 * 3) + (4

and then we add the missing parentheses and it becomes

    (1) + (2 * 3) + (4)

which seems to achieve a similar goal and is pretty unambiguous.

Recently seeing lots of paywalled articles on HN that quickly get tens of upvotes before they get any comments.

Wow, this works for any number of colors.

> Lost, are the "cut them slack" days.

I guess from the inside it feels different: I'll read 99 mind-numbingly bad comments and cut them all slack (in the sense of not replying to them at all), but these 99 instances of benevolence are invisible and count for nothing, because the 100th comment will make me fly in a rage and that's when I'll actually post something. And unload a bunch of my frustrations from the previous 99, too. The internet selects for extreme reactions.

That's a valiant attempt at myth-fighting, but it doesn't fully convince me. For example, one hop to Wikipedia gives this:

> Later, in the 1970s, ARPA did emphasize the goal of "command and control". According to Stephen J. Lukasik, who was deputy director (1967–1970) and Director of DARPA (1970–1975):

> "The goal was to exploit new computer technologies to meet the needs of military command and control against nuclear threats, achieve survivable control of US nuclear forces, and improve military tactical and management decision making."

Altruist? DARPA is a military agency, ARPANET was a prototype network designed to survive a nuclear strike. I think the grandparent comment's point is that the innovation was government-funded and made available openly; none of which depends on the slightest on its being altruist.

Wow. This reminds me of Google Sheets formulas, where function parameters are separated with , or ; depending on locale.