digging

Comparing "bug flying"/"bug survival" to "memorizing tokens" is disingenuous. They're not in the same category of task at all. You're essentially comparing the output of one system to the input of another system.

Ok so just to be clear, that's an entirely different and unrelated argument from the one I responded to.

Second, it's wrong. LLMs can learn within their context window. The main issue now is the limited size of their context window; animals have a lifetime of compressed context and LLMs only have approximately one conversation.

> In 1910 there were 240,000 carriage manufactures in the U.S, only 4 years later, there were 5.

This makes no sense. Carriages weren't invented in the 1900s. They were displaced by a new technology. LLMs are not the carriages of our world, they are the cars.

> We have no use for crypto-currency

You have no use for it, but a lot of people do (not just for speculation).

Getting bored of it isn't the same as having no use for it. This stuff is 2 years old; we haven't figured it all out. Even if all development on the actual models stopped now, we'd still be seeing new uses of LLMs emerging for years to come.

They lived so long ago, my understanding is such artifacts would be extremely unlikely to survive. We also haven't looked in that many places at that depth.

I mean it's still quite unlikely though.

> I think calling it intelligent is being extremely generous ... can't properly recognize that GitHub is capitalized.

Wouldn't this make chimpanzees and ravens and dolphins unintelligent too? You're asking it to do a task that's (mostly) easy for humans. It's not a human though. It's an alien intelligence which "thinks" in our language, but not in the same way we do.

If they could, specialized AI might think we're unintelligent based on how often we fail, even with advanced tools, pattern matching tasks that are trivial for them. Would you say they're right to feel that way?

Seems likely. That exact setup is also a (mediocre) fruit fly trap.

That's... a much less clever sentiment than it thinks.

> No matter what you say, you will look like a fool

That's because the question is nonsense, not because it's about the future. If the question is "What is the most powerful computer that will be built this decade?" it's answerable. No answer will be correct but there are certainly ways to answer that aren't foolish at all.

I don't think you've interpreted that quite right. Modern wars aren't funded by the bank accounts of the people profiting from them. War is a wealth transfer to the top.

Just as you say, it's a matter of it just being comparatively very easy to make games. These questions have easy answers:

> Like who made too many games?

Everybody.

> Who should stop making games?

Nobody.

It's simply that more games get made than anyone can possibly consume. That counts as "too many [to consume]". That doesn't mean anything should change.

It only feels weird because we recently came from a time where it felt like you could play every game coming out (except if you allow flash games, that actually hasn't been true for a longer time than it seems).

But nobody is complaining that "we made too many t-shirt designs in 2024 and I can't wear them all."

> To my mind, calling them Greek is a bit like calling people in, say, Belize "English".

Those situations are nothing alike. The Byzantines lived in Greece. Byzantium was founded by Greeks. Others in Europe called them "the Greeks". They were the genuine continuation of Hellenic culture for over 10 centuries.

It's either that or Greeks ceased to exist between Roman conquest and Ottoman independence - at which point they were ruled by a German and, presumably by your logic, were actually Turks anyway, not Greeks.

If you know much about Hellenic history, you know it's been a culture in flux since prehistory. I'd assert there has never been a group that you would call "true Greeks"... except maybe the Graecoi - Hellenic colonists in Italy. Even the post-Classical period of pan-Hellenism was driven and ruled by Macedonians, who a century prior were not considered Hellenes.

If they're Byzantine, calling them "Greek" is not wrong.

Byzantine = specific, inaccurate. Romans = accurate, nonspecific. Greek = a bit of both

> some people dying early while everyone else went about their business

So, evolution - just framed correctly.

> They are simply following the market.

This is badly misinformed. They make the market. When you buy a car, do you have it custom made, or do you select one that actually exists?

That's exactly what happened with SUVs. SUVs didn't happen because people were begging car companies to make them something big, dangerous, and wasteful. They happened because car companies in the US found a legal loophole where they could cut costs by skirting safety and emissions regulations, while simultaneously marking up the product as a premium one. Then they ran ads to tell people SUVs were super safe [for the passengers]. So when people started buying SUVs en masse, that wasn't organic demand, it was the result of a successful national misinformation campaign (because modern SUVs and other "light trucks" are so large you're more likely to just drive over your own child without seeing them than even get in a crash).

When you put it that way, the answer seems obvious. You're not paying taxes to the country you're residing in. You're not paying taxes for the infrastructure you're using.

Why is it ok for pure tourism? Because tourism is expected to be shorter-term, and you're likely to be putting more money into the local economy as a tourist.

So they need to register this at the very least. I don't know if they tax digital nomad work but they do obviously want to have some control over it.

I feel like everyone is missing "in 2011".

> As a resident of Oakland, California, I find this kind of technology a decent middle-ground. I'd like to be able to catch a decent percentage of criminals so as to make crime statistically pointless.

That's not at all what's going to happen. I don't think we have reason to believe it's going to significantly reduce crime let alone make it "statistically pointless."

What it will do is, you know, conduct mass surveillance. This technology will support the policing doctrine that everybody is a criminal suspect. That means you are a criminal suspect, and the police will watch you, and if they want to, they'll stop or arrest you for something or other. You don't need to be dangerous to anyone, that's irrelevant. Just vulnerable. And this tech makes you more vulnerable.

Is that a price worth paying for a pipe dream of safety?

The results of this election will not be knowable the night of the election anyway. There are way too many people currently preparing too many different ways to challenge the results regardless of what actually happens.

> Versions change: UI elements move, change their title slightly

Not randomly, I'd hope. I think you may be misunderstanding what deterministic means - or I am.

They're not going to try to do the job, they're going to hire cheaper, worse SWEs to manipulate AI... and then things will come apart pretty quickly :) But they'll still have someone else to blame.

> LLMs are a tool, nothing more, they don't magically imbue the user with competency.

Not a good take though, IMO. They're literally a tool that can teach you how to use them, or anything else.

I read it as "attributed to people who ... are generated by artificial intelligence.'

Insurance against the argument that "This person who wrote the review does exist, just not in a flesh body, they're an AI creation." But that might also be an instant-flop argument legally since I'm sure "personhood" has some definition near-future AI can't hope to approach.

I think we're largely in agreement then, actually. I'm seeing "world models" as a spectrum. World models aren't even consistent among adult humans. I claim LLMs are moving up that ladder, and whether or not they've crosses a threshold into "real" world models I do not actually claim to know. Of course I also agree that it's very possible, maybe even likely, that LLMs aren't able to cross that threshold.

> this claim ... is hard to evaluate without a well-formed definition of what it means to have a world model

Absolutely yes, but that only makes it more imperative that we're analyzing things critically, rigorously, and honestly. Again you and I may be on the same side here. Mainly my point was that asserting the intrinsic non-intelligence of LLMs is a very bad take, as it's not supported by evidence and, if anything, it contradicts some (admittedly very difficult to parse) evidence we do have that LLMs might be able to develop a general capability for constructing mental models of the world.

Well I actually read Neel Nanda's writings on it which acknowledge weaknesses and potential gaps. Because I'm not qualified to judge it myself.

But that's hardly the point. The question is whether or not "general intelligence" is an emergent property from stupider processes, and my view is "Yes, almost certainly, isn't that the most likely explanation for our own intelligence?" If it is, and we keep seeing LLMs building more robust approximations of real world models, it's pretty insane to say "No, there is without doubt a wall we're going to hit. It's invisible but I know it's there."

Actually, that's a quite good analogy. It's just weird how prolific the view is in my circles compared to climate-change denial. I suppose I'm really writing for lurkers though, not for the people I'm responding to.

> If you have nothing to contribute to the meeting, and nothing you need to hear from the meeting, just don't go. Unless you're working for one of those companies who care about meeting attendance from uninvolved people (which companies are these so I can avoid them?), then nobody is going to notice or care that you're not there.

You've overlooked my prior point: I'm not given enough information (any, actually) to know this beforehand. I can assume, because it's very likely true, but then once or twice a month I'm missing from a meeting where my presence is actually useful.

In an idealized training video, that's the moment where my manager would ask, "Hey, I noticed you've missed some important meetings, is there anything I can do?" and I'd explain the situation.

In reality, at the start of the meeting they'll bug me 0 or more times until they give up (regardless of whether or not I'm needed at the meeting, because they probably don't know either), they'll never ask what's going on, and their mental model will slowly adjust to see me as more and more "unreliable and disinterested".

> pure probabilistic sampling of words that anybody with a half a brain who doesn't drink the Kool-Aid can easily identify.

Your confidence is inspiring!

I'm just a moron, a true dimwit. I can't understand how strictly non-intelligent functions like word prediction can appear to develop a world model, a la the Othello Paper[0]. Obviously, it's not possible that intelligence emerges from non-intelligent processes. Our brains, as we all know, are formed around a kernel of true intelligence.

Could you possibly spare the time to explain this phenomenon to me?

[0] https://thegradient.pub/othello/

I do not think this is a case of that happening like with other big tech names treating customers like shit until they make it onto HN. Vlad's super active (like, is he ok?) on the Orion browser bugs/feedback forums so I have to assume the same is true of Kagi, their actual money maker. Sometimes things fall through the cracks. One data point isn't a trend.

I don't feel I have enough ground truth to know how good Kagi's search is in absolute terms (eg. there are plenty of searches where I still just "fail to find" and I don't know if it's my fault or theirs), I just know that I get less "junk" results with it than with DDG by far. With the additional ability to customize results and filters, I'd say it's a good product and it's worth a small subscription fee.

The quality of all free search is just bad. Kagi, even when it fails, is basically the difference between me looking for something or just giving up and deciding it's not worthwhile. Kagi's not magic-tier like early Google was - it's basically just the only modern web search engine.

As a big advocate of remote work, over the years I'm coming to agree with this less and less. Done well, remote work is great. Done poorly, it's killing me. It often saps me of energy even more than office work did somehow.

I have a strong inverse relationship between talking and fatigue in meetings. If I'm talking, it's generally because there's something for me to say, which means I'm engaged. I guess it does hit a limit where I've talked too much, but in my role that's pretty rare.

What gives me fatigue is the opposite. It's the meetings where I have nothing to say whatsoever. It's driven by the refusal, or inability, of organizers to plan or focus their meetings. I spend >95% of my meeting time trying to judge how little attention I'm allowed to pay to what's being discussed. Because I usually can't just leave and actually focus on work, but I also can't pay attention because the conversation is meaningless to me (think: discussing extremely specific problems of projects I've never heard of). There's no winning move, so "try to keep an ear open for my name or other keywords while I fidget, draw, or otherwise waste time" is unfortunately the optimal strategy in most meetings, even though it's extremely tiring. Can't imagine if I had to have my camera on.

Now that I've written all this out... I might just start leaving meetings and hope I don't get fired.