Brainwavez
23d ago
Stripping recently-granted citizenship is different, almost equivalent to revoking a green card, and maybe doesn’t need to go through the same justice system. Although if someone has been a citizen for a long time (say, around a decade), at that point they may have more ties in their host country: in extreme cases, if they moved as a child with no relatives, their birth country would be like a truly foreign country to them.
An immigration system should only grant citizenship to someone after they’ve stayed in the country for a long time, so every case of removing it should be the latter. But if a party grants citizenship to visa holders who’ve in some cases only stayed months (which may have already happened), the next party should be able to revoke it almost as easily as they could revoke the visas.
Brainwavez
23d ago
This isn't the solution, though, is it?
The academics only had a solution to a subset of Navier-Stokes, like that group. OpenAI seems to be the only ones who completed the proof, although Tao suggested the academics' solution may have extended to it given a bit more time.
The more important underlying issue is: frontier models are outcompeting aspiring academics in theoretical math, making open problems trivial, making them feel obsolete.
I think there's a workaround that lets academics (including Buckmaster et al) keep their pride and purpose.
Terrace Tao wrote a few days ago: AI largely autonomously solving open problems is bad, because the solution itself isn't very important, but the techniques (advances, insights, etc.) developed to find the solution and failed attempts are. These techniques assist future problems including ones with practical applications. Meanwhile LLM solutions are barely readable so we don't get any techniques from them, only the open problem solution.
So even if a model solves an open problem, it still makes sense for academics to do themselves, or at least break the model's solution into something comprehensible by humans, and receive credit for that. Buckmaster et al were already doing that:
I can say the first LLM generated proof Levent sent me was the most horrendous I have ever read; we verified it on Lean on August 22nd. Since this point, we have been working around the clock to understand this proof and turn it into something readable
So why don't they just finish their readable proof, publish, and get credit for that? Look, even if OpenAI hesitated to publish their allegedly stolen result, they still wouldn't be first, because another independent group has now also reached the solution. It's unfortunate that sometimes mathematicians reach the same discovery independently in parallel, but these ones still have a chance to be famous in a small group, just not in a (slightly larger?) small group.
Brainwavez
23d ago
Stripping citizenship is equivalent to criminal punishment: it takes away someone's rights, and has secondary consequences (loss of suffrage, property, community, in some cases effectively a death sentence).
Which means it should be fine if someone is tried in a legal system and could otherwise be sentenced to death, i.e. have demonstrated they can't follow the bare minimum, repeatedly breaking rules with no excuse. But if they're not tried at all or deported for a vague reason like "not assimilating" (where assimilation is defined beyond laws, or clearly defined but selectively enforced) - imagine if such process instead led to their fine, imprisonment, or execution.
In direct democracy this is true (that the majority is to blame for any mistake, so no one is really held responsible). But in representative democracy constituents will blame their representatives, even if they attempt every promised policy, for not implementing them correctly (or trying hard enough, etc.).
And in the US nearly everyone blames the president for everything, which is worse than a dictatorship, because even when he’s overruled, they assume he could just try harder and blame him for not doing so.
I know the list and bingo card are mostly rationalist science fiction, but I still think they would've benefited from Peter Watts, Issac Asimov, Ursula K. Le Guin, Greg Egan, etc. And with a measly 182 works, surely the aspiring reader has time for more.
/images/17886402945577037.webp
EDIT: Fargo, I watched the 1996 film, not whatever came out in 2015, but I can't change the image so it stays.
EDIT2: it's linked in the parent, but I also recommend the list creator's Substack, where he reviews many more works, of more diverse genres and mediums (e.g. Obsession, Blue Prince, Claude Plays Pokemon (a stream of Claude 3.7 playing Pokemon), Anthology of the Killer (underrated weird walking simulator)).
I liked Endacopia but less than I thought I would. Harsh but important criticism: I think it suffers from performative weird (likewise ENA, which Andy alluded he wants to be different from, and mostly is except this aspect):
I believe in two different kinds of weird. "Performative weird" is basically random: doing anything except what's normal and expected. There's no soul or deep meaning behind it, it's just weird to be labeled weird, performative. Then there's "true weird", where the artist genuinely tries to express themselves, not being weird on purpose, but not censoring themselves to conform to norms or expectations. The best art is true weird, I think (because) it's often more normal than society; in the sense that, society has lots of taboos that exist because of peer pressure, but don't fit reality or our internal values/emotions (yes, I'm talking about this site), and true weird lets the audience see through them. Even if I don't hold the same opinions as a true weirdo (which is almost always the case), it still illuminates reality and my values/emotions, by showing me novel ideas, letting me mix and match to form better ideas of those.
...anyways, I did overall like Endacopia, I do think there are parts that are true weird, and maybe I call the other parts performative only because they don't resonate. For example, I think the characters aren't very interesting or endearing (with exceptions, most notably Mellow). But I really liked Andy's signature, the pixelated warm but uncanny classroom aesthetic (for example, the blue room with the red floor and clock).
I also highly recommend its evil twin, Perfect Tides: Station to Station
The idea I hear is clear and simple: the techno-fascist billionaires will use AI, Palantir, drones, Flock, … to turn the world into 1984. It will be racist and sexist to show off power and control, and because the dictators are “right wing” so they have to do everything the “left wing” hates. Ordinary tech employees don’t want this and wouldn’t benefit from it, but they don’t think about it, trudging along on isolated tasks to collect a decent paycheck (The Banality of Evil).
If they limited their ideas to Mars or some bought-out city, maybe the author is motte-and-baileying, but I see no issue. Voluntary experimentation is good, maybe they do have some improved ways of running things.
There’s something to be said about how libertarians become fascists, and cyberlibertarians become cyber-fascists. I don’t think it’s a solution: tyranny collapses on itself, as the “victors” fight and die out, and doesn’t seem interesting (of course, also the ethical issues).
Brainwavez
28d ago
being able to learn after pre-training
True AGI must be able to learn continuously, or at least emulate it well enough.
Does this mean AGI is here?
AGI is here when we can't make ARC-AGI 4 (that a median human can pass but the frontier can't).
Bernie Sanders today has introduced legislation calling for a ban on superintelligence with a 20 year prison term for attempted superintelligence
He should've just cited AI 2040 instead of that vague proposal. The rationalists are genuinely smart and dedicating their lives to this, and I really believe they are the ones who are best equipped to handle regulation, even if their plans sound infeasible and sometimes it seems they've given up (and the other times they're naively optimistic up to delirium). Imagine what our pre-ASI utopia would've been if only politicians listened to autists.
I'm interested to see if it can write like a normal human and be interesting. If not, maybe it's for the best.
Ironically, I feel like this kind of proposal has a chance of succeeding, but only if it's guided by AI itself.
I would not trust a group of humans to not rig a datacenter or build one in secret, even "rigging" it accidentally (oops, it caught fire and our failsafe to blow up the other datacenter was never tested) or the secret datacenter not affiliated with the government but a rouge megacorp (but when it's discovered, the government turn a blind eye, because what would making waves accomplish for them?)
But a superintelligence, now that has the competence to carry out a Rube Goldberg plan without messing up a single detail. Just as long as we solve the hallucination problem...
Pre scientific people were intentionally mislead by smarter ones too: mystics, false religions, etc. And science is sometimes like mystic religion, with incorrect beliefs perpetuated by dogma, but I think it’s more accurate than those of the past: exposed papers do get retracted, my albeit basic understanding of mysticism is that there isn’t really an analogue “oh our beliefs are wrong now that outsiders have actually questioned them”.
Brainwavez
29d ago
I’m sure it has been mentioned before, but Anthropic and OpenAI benefit merely from advocating for a pause: it gives the impression their models are powerful and they care about safety (at a cost they persuade the government to regulate against their wishes).
See the Covert AI Projects supplement.
They acknowledge detecting a “small” training is impossible, but today’s frontier models need lots of GPUs which need lots of factories to manufacture, they presume even a nation state would have trouble concealing those without raising suspicion.
(FYI I think it’s more likely China would rig the Canadian data center and steal/defuse the Mongolian one, or the US would vice versa. I haven’t read the full document, maybe they address this too. I do credit these guys for covering their bases.)
This is something I thought about, but how do they expect to prevent China from pretending to cooperate but continuing in secret?
Brainwavez
30d ago
As flawed as academia is, I still think we'd be stupider without it. For example, imagine if we didn't have the psychological fallacies (although even today people fall for them way too frequently, imagine if you couldn't even point them out because they weren't established, and AFAIK these haven't been debunked).
Brainwavez
30d ago
I don't know about electrical engineering, but at least in math, I expect it's far easier to change someone's mind with evidence than it is in a social science. Even with informal proofs, there's a greater shared understanding and trust of axioms that, when put together, invalidate (or confirm) an argument.
Occasionally a math paper gets retracted for errors, but it seems far rarer than psychology, and with much less controversy. Meanwhile, many theorems are getting formalized in Lean, effectively "replicating" the proofs, many are successful and the vast majority are successful or pending.
Brainwavez
30d ago
I'm starting to think that Futurama was rather prescient in calling the turn of the century from 20th to 21st as the Stupid Ages.
No matter how stupid people are today, I'm sure they were dumber pre-1900, especially before we normalized hand-washing. Although if you want to call the entire history of humanity "the stupid ages" I'd agree to that.
what if it turns out even very "hard" or concrete fields like electrical engineering or chemistry are built on a bunch of stupidity that has just enough facade to be convincing to the in-field expert and certainly more than enough to the layman?
I'd be surprised if these didn't have subtle errors that will be corrected later, but they (even the slightly incorrect old models) have already been replicated and extrapolated into new discoveries, unlike much of psychology, so I wouldn't call them "stupid" and certainly not useless.
Brainwavez
1mo ago
Does Codex beat Claude at coding? Last time I used Codex it too much longer and over-engineered more than Claude, although that was a while ago.
To me, the reactions against 9/11, 10/7, and school shootings aren't much different. None are inevitable in theory (and certainly all are tragic), but in practice the optimal amount is not zero; there are responses that go too far, which indirectly kill more people while not even helping (especially in the long run), and people with hidden incentives abuse emotion to argue such policies.
But I agree with you that the counter isn't to dismiss 9/11 as not a terrorist attack or Hamas as not evil. We can still try to solve these problems in ways that improve society overall, and maybe reduce them, maybe in the distant future almost entirely; I suspect many people in the 1900s would be shocked at today's murder and starvation rates, falsely believing higher ones would always be inevitable. It's like how when people see stereotype-reinforcing statistics, they should understand that the statistic does not imply they should assume members follow the stereotype (e.g. inverse fallacy), but some dismiss the statistic entirely to reach the same conclusion, they can't believe both for some reason.
Brainwavez
1mo ago
What I don't understand (well, kind of) is why the "left" does so much damage here, but not to, say, GrubHub, whose pandemic ad I feel is 20x more offensive, insensitive to the most discriminated demographic.
country club golf gentleman (fine member of high society)
I'm just alluding to how country club members pride themselves on being civilized and "cultured", in contrast to the barbaric violent poor, yet here we purportedly have one assaulting a woman because she reaches for his golf club. I don't have anything against the country club culture, but I think they regard themselves too highly, believing their exclusivity and luxury improves their character, while they ultimately pursue shiny things and creature comforts like the rest of us.
Brainwavez
1mo ago
Static advertisements aren’t the worst (unlike animated) and may be better than nothing. But better than advertisements are murals commissioned to real artists whose goal isn’t to sell something to the lowest common denominator. I prefer those on walls and structures and imagine Fenway Park would look better with them too.
The latest stupid culture war news: The notorious Good Good golf advertisement. Featuring a country club golf gentleman (fine member of high society) having a manic episode, so excited to try overpriced "luxury" slop, he shoves a lady to the ground. Instead of calling the police, the other players just look on; maybe he's the owner, you know the type, with so much money and so little financial or social awareness that he'd even waste it on, random example, a "Quantum Triple Diamond Driver".
I don't fault lefties for boycotting Good Good, coercing them to retract the ad and apologize. Because they should do it to every commercial: I want every non-monopoly with aggressive advertising to be given this (or the Bud Light post trans ad) treatment, until public and online spaces look even better than São Paulo, Lancy Geneva, and Maine (where billboards are outlawed). At least this ad tries to not be boring and obnoxious; I actually think it's somewhere in the top 1% I see and hear (although not a compliment), thankfully rarely anymore thanks to uBlock Origin, SponsorBlock, and generally avoiding software whose primary function is ad delivery.
And this includes invasive tracking. Recently Twitter sent C&Ds to Nitter and XCancel. This follows Reddit tightening old.reddit and Facebook/Instagram have always been locked down. Thankfully there are always alternatives, right now https://twitterwebviewer.com https://nitter.kareem.one for Twitter, https://redlib.catsarch.com for Reddit, and https://inviziogram.com/ for Instagram - but it's annoying and contributes to link rot. This is your regularly-scheduled reminder to quit Twitter and the other brainrot social media, and replace them with more productive activities, like maybe, partaking in these "sophisticated" forum discussions on theMotte.
Brainwavez
1mo ago
It’s a fuzzy line.
What if the person isn’t baking a cake but officiating weddings, should they be forced to accommodate gay ones as long as they follow the same script as straight ones?
On the other hand, the person baking cakes won’t know whether their clients are gay without obnoxious vetting which would hurt their business. A man and woman can walk into the store and order a cake, then give it to their gay friends.
I believe freedom of association should work like that: a seller can refuse “service”, but once they provide it, can’t restrict the buyer or force them to retract.
I’m with the lefties that working hours should decrease to start matching increased productivity, with less demand for more luxuries vs more free time, especially with robotics improving and promising to further improve productivity in blue-collar jobs.
More options
Context Copy link