ben_w

> Either they're rational adults or they're not.

It's not a boolean; if you try to make it one, it is quickly apparent that there's no such thing as a totally rational human being. And I don't just mean amongst the electorate, I mean you won't even find one working as a professor of rational logic in a university of mathematics and philosophy.

> If you think voters’ minds are so feeble that they can be swayed by propaganda, then the problem is the voters, not the propaganda. Anyone so easily manipulated can’t be trusted to decide who is in power.

Propaganda works even without free votes, this is why dictatorships still bother with it.

Unfortunately this is a non-trivial problem; I don't know if Hacker News has a size limit on comments, but whole books can be (and have been) written on this topic and its implications for democracies, more for the implications for dictatorships, and yet more for the implications for free trade and capitalism.

Same reason they take private investments or loans: To grow faster than would be possible without that.

On the one hand, he said he'd go to jail if the dems won. On the other, he says a lot of things that aren't true.

At this point, Musk might face extradition claims outside the US and state-level charges within it, even if Trump pardons him for all US federal laws. And he might have lost a lot of leverage with the US federal government by having SpaceX stop taking customers for more Falcon launches before Starship's fully ready.

> It would be unethical to even attempt that.

Then pick a smaller organic brain which doesn't have the ethical problem.

The closest we would know about how to teach a honey bee calculus, is to train an LLM, then stick around a billion tiny electrodes into the honey bee and do Hebbian learning to rewire it completely to match the LLM, which would likely completely erase anything the honey bee had learned, or it might not because we are missing something fundamental with Hebbian theory and (AFAICT, I'm not a neuroscientist) we don't all agree on what.

For me, DIS wasn't so much a case of "the problem with" as "amongst the many problems with".

Things as you say, but also: didn't know show's own technology history (Klingons with cloaks before they should've had them), adding in spore drive then needing to retcon it.

The show din't seem to know what it wanted to be and kept pulling in different directions, one of the early plans was an "anthology series, with each season being a standalone, serialized story set in a different era", which would have suited this much, much better: one or two episodes about the Klingon war, one showing Pike's command before Enterprise, young Spock's first mission, even some about a spore drive and the problems it has, some AI-takeover stuff, some post-TNG stuff, one episode about galaxy-threatening event, one about some post-apocalyptic thing like The Burn isn't even bad. But all that smooshed together with the same crew? Nah.

The reason Michael Burnham had a male-coded name was teased but never actually revealed (could have been interesting e.g. "adopted by Vulcans who didn't notice it was male-coded because that's a silly human thing"). Her character shifted way too hard and way too fast, from "Vulcan persona due to upbringing" to "traumatised normal human", a nature-vs-nurture tale could have been an interesting thing to explore.

And Michael Burnham was correct to want a show of force against the Klingons! Sure, wildly violated how command hierarchy works, but she was correct, and Starfleet Command should have known that because she got advice from Sarek. Not that she should have been Sarek's never-before-mentioned adopted daughter, we need to see new characters rather than just repeating the old ones; I mean, Vulcan is an entire plant, not a small Pacific island. (Amusing trivia from some fan-site I visited decades ago: VOY had the opposite problem, in that the list of extras appearing as background crew was larger than the crew complement).

And in the battle before they went off to the far future, that unexploded torpedo inside Enterprise, which the admiral heroically sacrificed herself to disarm… (1) it's antimatter, how does that "not explode"?, (2) it's ~ 50 megatons and inside the ship and in TOS an overloaded phaser pistol was supposed to be able to blow out multiple decks, (3) what, are their transporters not working? Can't you either beam it out, or the admiral out after she disarms it? (4) Why send in an admiral, did they run out of ensigns and non-officers? And Control (the AI) was proto-Borg, right as we had three seasons of Picard with season 1: Romulans, Borg, and Androids; Season 2: Q, Borg, and time travel; Season 3: Borg, Changelings, and Crusher saying "no one has seen or heard from the Borg in a decade".

> I have a deep affection for stories where people are depicted as good, or even just good at their jobs. I appreciate for actors and writers it's a much harder ask - it's much easier to make a story of relatable bumbleheads, especially if you yourself are somewhat uninteresting and never got much life experience. It's much harder making believable characters that we can look up to and respect.

Mmm. This represents what I liked about the TOS-ENT era of Trek, SG-1, Babylon 5, Lower Decks, and roughly half of SNW; and also what I dislike about DIS (the nice crew are noobs, the competent ones are dicks), season one of PIC, half of what I don't like with season 4 of SNW, and why I was so bored by Stargate Universe.

> EROEI

Modern PV typically has an EROEI of ~ 15-30, depending on tech, location, system lifetime, and accounting methodology.

From what I've read, coal might still beat that, but modern oil no longer does. Gas is generally also worse than PV in this regard.

Coal is however also very expensive, which is why e.g. the UK completely stopped using it for electricity.

"Cheap enough" doesn't exist as an isolated term, it's "cheap enough given task involves some specific competence".

I can vibe code a card game on an unpaid Claude account, so it's definitely cheap enough to replace developers (and businesses selling card games) if you just want a card game.

A whole modern operating system? Right now, I think you'd still need to be a developer just to direct the AI well enough to make more than a demo.

> An AI trained on leaked Apple ROM code?

Nah, just one which read e.g. every Hacker News discussion like this.

> What you do on your own machine is your business, so have fun with that. You can't legally share it, though.

Unclear, as with so much else going on in this space right now.

Don't get me wrong, it feels like it ought to be against the rules to me… but then again, several highly successful businesses also felt to me like they ought to be against the rules.

Hm. I think it depends on two things: (1) how sample-efficient AI can become if CS and mathematics becomes "solved", (2) if AI still sucks on this metric compared to biological brains despite that, can it at least self-play its way to solutions?

Given what AI could do with protein folding, I wouldn't bet on biology being a "no" for AI. It might be, but I wouldn't bet on it.

Likewise, if AI can do mathematics so well it solves "just" all of computer science, let alone other branches of mathematics, if this somehow doesn't lead to AI becoming similarly sample-efficient as biological brains, that's very interesting and will say something nice and surprising about the nature of minds and possibly the nature of being alive.

Good for you. But Haiku (OS) is MIT licensed and therefore easily forked, and at this point the AI you dismiss as "Special Needs Wintermute" (much as I agree with the sentiment "You need to be able to explain your work without just pasting in what [it] says") is being used to find and resolve critical issues in OSes that actually get used outside of hobbies, and even when the AI was a lot more meh than it currently is, it has been useful for such hobby projects* when used by people capable of explaining as you demand.

* I like hobby projects, no shade here. But it is a hobby project, just like the people trying to keep RISC OS modern, or my suggestion about MacOS classic. There's no obvious reason to pick these things over "just use some flavour of linux" in the present world, unless you're a nerd like us.

Ironically, it being a hobby project might actually be a good reason to keep the AI away: the effort is the point.

Several AI companies are getting sued for what their products allow users to do, as was Google for what YouTube allowed its users to do.

YouTube is probably an instructive example here: they can refuse to host certain content on their servers, and this is Anthropic refusing to have the AI on their servers perform certain tasks.

> An alternative approach is high-level emulation. Running a separate instance per application gives you process isolation (via the host OS) for free.

The AI know this too, and that is a way to rework the OS.

> It had most if not all of that.

I could have phrased this better, the "modern" is a prefix on each item in the list:

  modern {memory protection, virtual memory, preemptive multitasking, and networking}.

rather than:

  {modern memory protection}, {virtual memory}, {preemptive multitasking}, and {networking}.

I thought it was the PPC chips, not the OS, which emulated a 68020?

Sounds about right. Similar experience learning German when I moved to Berlin from the UK (though less extreme, felt like 85% of words was 50% of sentences).

Mmm. Or improve the OSes themselves.

At this point, they could likely decompile MacOS classic and rework it to give it modern memory protection, virtual memory, preemptive multitasking, and networking.

I assume booting an OS with a mere few megabytes of RAM footprint would only take milliseconds?

The description of vibeOS in that video is half way between what I thought I was replying to and what I suggested.

Doing it in a way that isn't described as "hilarious" is probably also possible in the time since that video came out. Doing it in a way which isn't described as "a security nightmare" or "liable to lose all your data", probably not yet. Not as vibe-coding, at least.

> I want to do MORE with the same amount of folks.

So do I, but I've been in two companies that had to let me go because the investors said they didn't want my role to exist any more. Plus two other roles where the company straight up ran out of money.

It's not always up to us.

> A while back I saw some people saying we would just have LMMs hallucinate an entire operating system every single frame. I guess technically it's now proven to be technically possible, but it would be thousands of times slower while requiring thousands of times more compute (a multi-million x difference from an ordinary OS!).

Gosh. That would be a very silly use of an LLM.

Bad use of any AI, even image/video models which are certainly less wrong than this. Stuff like DLSS show we're getting stuff that could generate a game's graphics frame-by-frame, but even though it's possible to have an entire game environment (not just graphics, everything) done by an AI, such things demo better than they play due to temporal coherency.

That said, having an AI make the code for each thing as you need it, might work somewhat OK eventually. Wouldn't trust it today (security, data integrity, and a tendency to cheat on tests), but eventually it could happen.

Edit: never mind my original, I looked at the about page:

  I studied law in Berlin (J.D.-equivalent) and at the Center for Transnational Legal Studies in London.

--

Original:

Hmm.

Would be interesting to hear from actual German lawyers about these examples; my understanding of law is (a) at the level of pub trivia, and (b) more about the British system, which is fundamentally different in any case*.

* https://upload.wikimedia.org/wikipedia/commons/9/92/Map_of_t...

> (This is a joke, to be clear.)

Good that you said so, this is annoyingly plausible.

This was true last year, but it would be risky to assume it remains true from one week to the next.

The rate of growth in this sector is "1. move so fast legislators can't keep up; 2. keep going until you hit a limit, then find a way to hack around that limit; 3. GOTO 2".

I have read that comments which appear to have been written by an LLM are auto-filtered. As these appear to mostly consist of quotations of LLM output, I expect this would be why.

I suggest putting them on a blog or something and linking to that, rather than HN comments.

> I wonder how this squares with the recent "NNs are already clean" interpretability finding, which may have also been the impetus for this kind of research. Similar to their work on CNNs, maybe residual-steering is overkill, there might be just a single neuron that "cleanly" corresponds to "evilness" that you can turn up/down?

By my reading of the TL;DR of "NNs are already clean" (given the full thing is a 65-minute read, I've not been through it yet), the result can be rephrased as "it made an ontology different to ours, so we were confused". (Example: we might expect a language model to identify nouns and there to be a noun-neuron, but the neurons actually trigger at punctuation based on the emotional cue from the preceding clause).

Regardless, a one-neuron evil switch is quite clean, possibly (I'm not sure) clean enough for the older processes to find it.

The conundrum: (1) this was only discoverable because of open weight models, and research labs (and governments, if they regulated faster than the glaciers are melting) could be informed by this; (2) yet because it is known, it now means that it's easy for mid-skilled people to download open models, flip the switch to make the model evil, and then setting them lose to do evil things.

Security by obscurity vs. fully automated insecurity. Clichés need not apply.

And all the 420/69 references. And doing "let that sink in" as a live-action bit.

Only if it has a goatee, which is how we know Robert Downey Jr's Iron Man and Benedict Cumberbatch's Doctor Strange were only the heroes of their own narratives, and nobody else's.

Similar, though for me LLMs have recently noticed I'm writing a book.

Weirdly, one of them "thinks" I'm a completely different person (by name) and CCed an old work email address of mine for a company that ceased to exist about 15 years ago. I'd expected them to mix me up with a different person of the same name, but no.

I think some sarcastic responses to Mendeleev at the time actually suggested this.