karahime

It's interesting to see the old "Why would we go to space when there are still uncured diseases" show up in a place like this. Science and discovery are singular, all discovery aids all discovery.

I think it's pretty justified. The internet was unfairly maligned as a place solely for weirdos, nerds, and the unhinged by a lot of media for a long time, and often still is, even as everyone is online. You're thinking of the fanfare, but the fanfare is what is being treated dismissively as "just hype" within this same comparison.

You fall back on agree to disagree, twice, here, when you're accusing me of thinking only of my own perspective. It's not courteous to agree with something for the sake of agreement, nor is misrepresenting what we can say from an experiment useful for "showing patterns". You can use analogies without making poorly constructed, sweepingly grandiose ones.

Beyond that, like any parent, you have a lot of latitude in how you choose to treat your children, but I'm not going to agree that isolating them from computers is helping just because you frame it as "gentle", and I'm not going to say that we should replicate that elsewhere.

Old style rat drug experiments with rats in a barren cage being administered morphine are barely a good behavioral model for rats, let alone people. This is the aforementioned moral panic language. If the point was that solid it wouldn't need to make this multi-step analogy.

No, they're not, and yes, it is. I sincerely hope we do not take phones and other computers from future generations. That would be a terrible thing to do to them.

If we expand the placeholder, I think "Anything that can be rewritten in a language with excellent performance and safety characteristics will be rewritten in a language with excellent performance and safety characteristics" is a desirable state of affairs.

I don't think it's a gotcha at all, they openly said it and were pondering how that might help them. Misinformation plays more to sides that are wrong about more things, and I don't consider misinformation my ally in anything.

You should consider why the best ally to your position is misinformation.

Nobody wants to look at it, but I think this is a lot of it. The original Prince of Persia PC/Amiga version retailed at $39.95, or roughly $100 in today's money. In fact, roughly $100 in 2026 dollars was actually quite a stable price point for a very long time.

A $60 game in 2006 is also roughly $100 in 2026 dollars. I think what happened is that around then, people generally decided that "$60 is the price of a AAA game, and anything more is a sham". If I had to bet, I'd guess that that was around when a lot of the public got into gaming as something that wasn't just "for nerds".

I think that if people were willing to drop $100 up front, base price for a game, you would not find the microtransaction, ultimate edition, day 1 DLC shenanigans that you see. Companies would find it easy to say "But of course all skins are unlockable in game!"

I agree with this, and would add that it's hard in part because it's always been hard, and people have overcome. I can only imagine the difficulty of coming up with something like HTTP, or a suspension bridge, or algebra, without the mental scaffolding already being there. If you go back and read the original discussions of these, they include a ton of what seems like circuitous explanation for what we take for granted as "simple" ideas, but which are absolutely not simple at all if you have to pull them from the primordial void sight unseen.

Which is exactly the problem with this whole discussion. On the far side, you hear that it's heroin! It's fentanyl! It's alcohol! Facebook groups are the modern opium den! But when actually challenged, it's oh no no, that's a metaphor, it's metaphorical fentanyl, not real fentanyl. People on Instagram are metaphorically injecting metaphorical drugs into their metaphorical veins.

It's a poor basis for policy and thought. I would wager 20 francs that none of these people have ever seen a heroin OD. The whole discussion centers around a maximally impactful comparison but the middle of the comparison is hollow.

On the world being different now, you know, the post tries to make the case that it's not, and that the situation was also like this before LLMs. It tries to diagnose some deep root cause.

What I'm saying, though, is that if this were a Python project that were rewritten in C++ for speed, I think all of the above would be wildly out of line for GvR to write a teardown about how he tried to tell them how to write fast Python, but they were just too VC brained to listen.

It's not about the tone of the blog post. As a language designer, you already have complete control over what you want your language to be able to express, in a very literal and direct way. If he didn't like the Zig code of Bun, maybe that suggests that he's not putting the things he actually wants to be in Zig in Zig, or hasn't defined the separation between the language and what is written in it cleanly enough. Or, you accept that that's the beauty of it, and that poorly written code that runs runs.

It's not that in the end he couldn't control what Bun does, it's that he couldn't and shouldn't have been able to control that in the beginning, but the post is acting like he should have been able to assume direct control anyway.

Isn't this what makes it so strange, though? You didn't see K&R publish "so-and-so writes bad C", or Stroustrup decrying the Boost maintainers as hacks. Linus used to do this sort of thing, but mostly to things that directly affected the kernel, and even that eventually led to changed behavior and a code of conduct.

The post disclaims the ambassador relationship, but treats Bun as having all of the responsibility of being an ambassador anyway. If Bun is in Zig's house, like the monthly meetings and the core team code review suggest, then somehow the outcome reflects on Zig. If not, then that's fine, but then it begins and ends at "They were a project written in Zig, and now they're not". He can't have both "It's their fault because their code was slop" and "it's not my business".

Most of the US's problems on this have little to do with putting too much priority on safety. There are countries that show that you can have a sane regulatory process and still get well priced nuclear. France from the 70s through the 90s and South Korea are the classic examples. Neither compromised on what's actually needed, but both cleared the way on redundant walls that don't demonstrate or stop anything bad from happening. You can achieve both greater nuclear safety and reduced process burden through standardization, which is how most places that have done it got it done.

No. GAAP already prevents what it is implying from occurring, and the website is aware enough of this to admit it in the footnotes, but tries its best to distract from that with big, spicy implications on top. GAAP revenue is booked when there are real deliverables at arm's length pricing, and isn't when there isn't.

I agree, and think the effects on learning should be doubly emphasized. One can lock down everything and everyone to the highest degree possible, think of every possible edge case, set controls 2, 3, 4, 10 steps away from them, but not only is this not beneficial to society overall due to how it hurts adjacent information, it's not even beneficial to the goal in question, since it creates a brittle situation with locks that can't be changed or updated in a world which is always changing and always updating.

Personally, I think understanding deeply how a transformer works helps a lot to understand what's probably the result of specific choices in the RL process vs what's architecture. A lot of the "We asked 30 LLMs and they all said the same thing" type analyses of how LLMs work often bump into what's being prioritized in the name of alignment right now, as opposed to architectural insights.

News is famously 24 hours, and will always tell you to tune in at 8/9/10 PM to find out why the country is in a shambles/your home is under attack/the aliens are coming to abduct you.

It's the nature of the safetyist position. It creates everything that it claims to avoid (duplicitous behavior, misaligned outputs, unresponsive systems).

Sure there are. Ai2's OLMo, EleutherAI's Pythia, and LLM360's Amber are actually open source, top to bottom, training data to checkpoints to code, to name three of them. You also don't have to look far on Huggingface to find smaller models developed with open source processes all the time.

I don't know of any modern workflows that rely on "we'll tell the person not to do it again", though. There's a reason that companies have adopted blameless postmortems, because if your response to the DB going down is "It's fine, Kevin learns and next time he won't misuse the prod credentials", you are guaranteeing prod will go down again in the same way at some point.

See, I have to have my driver's license, but if I could have that on my phone as well, I might do this. Running out of battery is largely not a concern for me as I already carry an external battery with multiple days of charge.

Yes, just like you can't eat trucks, roads, grocery stores, tractors, combines, crop dusters, grain silos, mills, or the FDA. Any system needs many components, most of which aren't directly consumed by someone.

The vast consumer adoption and ongoing involvement seems to point the other way, though. I think a lot of the appearance of backlash is on (specifically anglophone, mostly) social media, which is going through a somewhat reactionary phase regardless.

You're mixing different things. Mobile first is integrated into new services to the point that they either are mobile first, or they have a design system which includes mobile as a surface. VR has a wide user base (MQ2 sold as well as the original Xbox) and is involved in both manufacturing design and simulation, and is hardly an academic taboo, even if the "main" topic of discussion is elsewhere right now. Blockchains are being integrated into financial infrastructure even as some people make snarky commentary about it. Sometimes optics is just an optical illusion.

Not likely. Take with whatever grain of salt you'd like, but that was largely a property of development being academicized and subject to things like grant cycles, research topic fashionability trends, and institutional structure. It would be wrong to assume it's some baked in thing that's guaranteed to happen independent of how development looks.

You're being sarcastic, but I do enjoy it. I just took a Waymo recently and it was thrilling, it felt great to feel the wind and the sun, listen to music, and get where I was going without having to drive there. I still like driving, obviously, but being able to decide one or the other is wonderful.

This has always been just a meme. Plenty of the people I know who know the most about tech are also very excited about it. Person B in this example is often miserable for a lot of other reasons that have nothing to do with tech, doesn't understand tech as deeply as the meme suggests, and pattern matches cynicism into everything.

This is an ongoing misunderstanding of the recent moves here. Willingness or unwillingness to spend on frontier models isn't derivable from whether cost caps exist. Comparing it to cloud spend, most companies have cap regimes to ensure that employees aren't spinning up the largest instances for every job and leaving them open, but that doesn't mean those instances are too expensive for what they do or that companies are rejecting them. Rather, it's an indicator of ecosystem integration and a common step in corporate policy structure around spend. The actual content of article talking about tier guidance and spending dashboards supports this reading more than the one implied by the headline, I believe.

This probably isn't totally untrue, but, for example, 4chan is sorted by recency and has a lot of the same problems as other large sites online, which to me says that it can't just be complex recommendation systems driving problems.