BANNED USER: temperamental dysregulation
DaseindustriesLtd
late version of a small language model
Tell me about it.
User ID: 745
Banned by: @Amadan
DaseindustriesLtd
late version of a small language model
3yr ago
Another reason is that local degenerates will simply rob you or set your property on fire or poison your livestock if you keep showing off with how trad and industrious and put-together you are. Many such cases with naive folks who tried returning to their farming roots. You can only do it with a reasonably large community that can afford security, and at that level you're perceived as a threat to the authorities or local mafia.
It's very much a shame.
DaseindustriesLtd
late version of a small language model
3yr ago
Your auth-right commune might get sent wholesale to siege Popasna or reinforce the line near Bakhmut or whatever. Alternatively, denazified via prison rapes or set on fire by Chechens. Also the climate is terrible. Other than that, a decent idea.
This isn't the first time Europeans bust a conspiracy with a surprisingly competent military prepping and an inane, politically doomed premise. I guess it's a package deal – if you really reject democracy, you cannot recognize how it recruits people to defend itself.
When I don't have energy for cooking and go to eat out, 9 out of 10 times it's a cafeteria with a small fixed menu. They serve a soup, a few bread buns, a fresh salad, and a dish with decent amount of chicken (12 options, realistically 5), all for something like $4,8. Right now I'm feeling cheeky so I'm in a mid-tier Iranian restaurant where Fesenjan and other delicious things can be had – for about 300% more. It leaves a better sensation after the fact, more giddy lightness than stuffed stomach; might just be saffron. Maybe I'll come here once more in my life.
Admittedly, this is unsolicited blogging and the moral of the story could as well be conveyed with McDonalds vs. fancy burgers, or with Scott's own musings on Coca-Cola in How The West HWas Won. I much prefer Wyclif's commentary. But there's no mystery in why Scott is a star of greater magnitude than Wyclif.
I've also wondered about prominence of rationalist thought leaders. But after all is said and done – this is cope, sorry. We know Scott's history and we know Eliezer's, it's all in the open, and many of us have been there for much, or all, of the way, long before they received any establishment attention. What makes those guys special is that they are, indeed, that good. Or rather – they're good at this niche public nerd-intellectual schtick. Good enough to have reasonably long shelf lives.
The first virtue of a public intellectual is productivity. The output must be consistent, voluminous, topical, wide-ranging, and ideally give the impression of effortlessness. A proper PI must be able to write about anything and, perhaps, write anything – from poetry to meta-analysis to an alchemical treatise; all in the same recognizable but immensely varied voice, a one-man magazine. It's a bit of a sports performance – namely, verbal gymnastics. Like Andrey Plakhov says:
…As those who download the pdf [of SSC blog] may notice, there are 9,500 pages. These are high-quality, well-written, thoughtful texts. To have prepared much of it, the author had to read dozens of articles, comprehend and conduct meaningful meta-analysis. Scott Alexander accomplished all of this in about seven years, while also managing to become and remain a practicing medical specialist (and this in the States, where the second task itself is often described as «the race of a lifetime»). Purely quantitatively, this is more than I thought possible for a human being. Personally, I'm not able to write at that rate even for days at a time, and even while on vacation (for clarity, let's drop the quality issue). But this is not something superhuman and does not make me suspect a «collective of authors». Rather, my feelings are similar to those of an amateur athlete who has watched how training of an Olympic champion looks like. Great motivation. [...]
Why am I all about numbers. Because we are faced with a case where it doesn't even translate into quality, but is quality in and of itself. On any topic, from the highly specialized to the generally important, Scott Alexander is able to fit a text of a couple of dozen pages long, stimulating further reflection simply by the mass of associations and unexpected angles that arise. Where I disagree with him, the disagreement is closer to «Wait, there might be another explanation,» or to «This line should have been thought through rather than brushed off,» rather than to «God, why am I reading this at all.»
Adding to that, Scott had his Annus Mirabilis of 2014, writing an entire book's worth of important articles! And they are all very accessible. That's the second virtue: clarity. Or put another way, being able to distinguish the job of a public intellectual and an academic performer. There is a niche for obscurantism, as Moldbug has demonstrated, but it is not large. People who'll read you are mostly not geniuses, but neither are they fools who can be intimidated and amazed by big words. Many of them just want an entertaining, clear narration and commentary of ideas they could as well write about themselves, had they more time.
Plakhov then proceeds to criticize Eliezer. I won't touch Eliezer beyond saying that he has essentially cultivated the community and intellectual culture where Scott got promoted, wrote an entire mini-Talmud of rationality and a very voluminous, popular work of fiction with the intention to peddle his ideas to the masses – an accomplishment that's harder than it might look. And this is the third virtue of a public intellectual – working in the context of a public-facing intellectual tradition, and molding it, and contributing to its growth. You can't get far on your own; nobody will care about a loquacious manifesto-shitting crank, even if by some miracle your manifestos are actually great so long as one looks past the inevitable lack of polish. So you've got to live in a society. It's not different from what aspiring artists (or sex workers) do, reposting and boosting each other. But these two have done more than others.
The fourth virtue is letting your personality shine through your texts - and better if it's a likeable personality for a wide enough audience. Author-clientele relationships are parasocial. Many very smart people would rather keep their identity small; so it takes reading 20x as much to get a good feeling for who Gwern is. It's perhaps a more authentic impression than one you get with a single note by Scott – but people like the latter more, and they like that it happens so naturally. But you mustn't be larger than life – that's the turf of life coaches, and it's nauseating for the target audience of intellectual commentary.
And only the fifth virtue, I'd say, is sticking to the confines of the Overton window, and thus preventing the loss of people who feel uncomfortable outside.
By the way, why do I know what Plakhov says? Because sometimes I get engrossed, perhaps infatuated, studying a hitherto ignored smartass, amazed that something so good, so obviously beyond my ken, was so readily available; and in a few days I skim his entire output, and in a few days+1 I am disappointed and see what should have been done and said better. But I did not. Because I'm not that productive and not that good.
There may be some factors loosely along the lines you hint at. Bluntly, they both are cult leader figures («rightful caliph» jokes are not just jokes), or perhaps intuitively exploiting the Learned Rabbi archetype of their ancestral culture. When I saw people debating Sequences IRL, it was hard to distinguish from a chavruta following a Shiur; or from an underground Marxist circle, acolytes diligently inspecting the complex wisdom in teacher's words. That's funny to see. That could contribute to the resiliency of their fame. That can be at most a part of it.
It may not be a perfect meritocracy. But it does select for the fittest.
DaseindustriesLtd
late version of a small language model
3yr ago
Marshall Brain is, from what I understand, a socialist, and thus given to conspiratorial reasoning where the class of actual, as-recognized-by-Marx capitalists are organically hostile to the rest of the society, more powerful and more connected. I think you raise valid objections against this viewpoint and its implications.
But it's a matter of degree, not kind. For those of us who are neither PMC nor large property owners, it mostly changes the schedule of going into terrafoam. The incentive of efficiency is the same.
Besides. In Manna, he says that this largely dehumanized welfare population is <230.600.000 people (<1500*cube capacity). The main character could work for 10 years after the major breakthrough in automation, and even had a stint as an administrator. So in our timeline that's 2032 at the earliest. Population Pyramid says there'll be 352 million Americans by that point. Census Bureau is more bullish and expects more like 360 million. So very naively, 360-230=130.
36-37% of the population is quite a lot more than Peter Thiel and Brussels class. We don't have that many NYT journalists either.
DaseindustriesLtd
late version of a small language model
3yr ago
similar to similar predictions of worsening regarding Pelosi's Taiwan visit ( remember that?)
Soon after that the US has obliterated Chinese semiconductor industry. But yeah, I rather expected some novel activity from the Chinese side.
DaseindustriesLtd
late version of a small language model
3yr ago
Saying it many time does not make it true. China’s military growth is described in the annual DOD reports to Congress. Most analysts regard it as pretty accurate and comprehensive. So references to “the rest of the iceberg” are mere hyperbole and political posture.
I've said it before, I'll say it again: Americans can't get their junk up without posturing as underdogs. There's the Sputnik moment, there's the missile gap, there's the supercomputer gap and AI papers gap and who knows what else... I guess this has been going on for so long, and so profitably, we're back to the missile gap. There are real gaps – in unlivable apartment buildings, raw concrete production, stuff like that – but they aren't discussed as often.
I suspect Americans have learned this despicable practice from their Israeli friends who are somewhat easier to emphasize with, but it doesn't reflect well on an uncontested hyperpower.
I'll believe the missile gap when I see it.
In any case, the age of the nuke is over, the fear is no longer there, and the West is getting ready for a conventional showdown, tightening the circle, increasing military budgets and ramping up production. This war is but a warmup.
DaseindustriesLtd
late version of a small language model
3yr ago
The appeal, from the perspective that's being talked about here, is deciding what kind of people (or some other agents, or infrastructure for some long-term purposes) to spend post-scarcity on, instead of sharing the Earth – nay, the entire light cone – with 9 billion unrelated poors. And if you don't share much, well, once again there's not really a lot of a point to them.
DaseindustriesLtd
late version of a small language model
3yr ago
Because cats are a poor fit for hotel lifestyle.
DaseindustriesLtd
late version of a small language model
3yr ago
Ironically @2rafa is a big fan of this particular technology, gait recognition, as a tool for suppressing violent crime. It's not like those programs and models are secret – any developed nation can do it.
But of course only rather intangible things such as «social contract» can stop the state from redefining most any dissent to crime.
In practice, if you run the numbers, UBI requires redistribution to increase substantially. ... Maybe the singularity brings such unimaginable abundance that money ceases to have any relevance, but even if it does there will still be a transition period.
A classic joke:
– Vodka has gone up in price, son.
– So Daddy, now you'll drink less?
– No son, but you'll have less to eat.
What I mean to say is that there are other solutions to the problem of unproductive masses, and they don't involve either outright genocide nor communism. Such as...
The government had finally figured out that giving choices to people on welfare was not such a great idea, and it was also expensive. Instead of giving people a welfare check, they started putting welfare recipients directly into government housing and serving them meals in a cafeteria. If the government could drive the cost of that housing and food down, it minimized the amount of money they had to spend per welfare recipient.
As the robots took over in the workplace, the number of welfare recipients grew rapidly. Manna replaced tens of millions of minimum wage workers with robots, and terrafoam housing became the warehouse of choice for them. Terrafoam buildings were not pretty, but they were incredibly inexpensive to build and were designed for maximum occupancy. They clustered the buildings on trash land well away from urban centers so no one had to look at them. It was a lot like an old-style college dorm. Each person got a 5 foot by 10 foot room with a bed and a TV — the world’s best pacifier. During the day the bed was a couch and people sat on the bedspread, which also served as a sheet and the blanket. At night the bed was a bed. When I arrived they had just started putting in bunk beds to double the number of people in each building. Burt was not excited to see me when I arrived — he had had a private room for 10 years, and my arrival was the end of that. At least he was polite about it.
... Downstairs there was the cafeteria staffed by robots. The robots were not bad — the food was acceptable. They also kept the bathrooms, hallways and rooms spotless. Every day at 7AM, 12 PM and 6 PM the breakfast, lunch and dinner meal shifts began. There were six 15-minute shifts per meal to save on cafeteria space. Burt and I had the third shift. You sat down, food was served, you ate, you talked for 5 minutes while you drank your “coffee” and you left so the next shift could come in. With 24,000 people coming in per shift, there was no time for standing in a cafeteria-style line. Everyone had an assigned seat, and an army of robots served you right at your table.
... Because no one had a window, they could really pack people into these buildings. Each terrafoam dorm building had a four-acre foot print. It was a perfect 417 foot by 417 foot by 417 foot solid brown cube. Each cube originally held exactly 76,800 people. Doubling this to 153,600 people in each building was unthinkable, but they were doing it anyway. On the other hand, you had to marvel at the efficiency. At that density, they could house every welfare recipient in the entire country in less than 1,500 of these buildings. By spacing the buildings 100 feet apart, they could house 200,000,000 people in a space of less than 20 square miles if they had wanted to. At that density, they could put everyone in the country without a job into a space less than five miles square in size, put a fence around it and forget about us. If they accidentally dropped a nuclear bomb or two on us, we would all be gone and they wouldn’t have to worry about us anymore.
.. Ultimately, you would expect that there would be riots across America. But the people could not riot. The terrorist scares at the beginning of the century had caused a number of important changes. Eventually, there were video security cameras and microphones covering and recording nearly every square inch of public space in America. There were taps on all phone conversations and Internet messages sniffing for terrorist clues. If anyone thought about starting a protest rally or a riot, or discussed any form of civil disobedience with anyone else, he was branded a terrorist and preemptively put in jail. Combine that with robotic security forces, and riots are impossible.
The only solution for most people, as they became unemployed, was government handouts. Terrafoam housing was what the government handed out.
DaseindustriesLtd
late version of a small language model
3yr ago
My family had a cat when I was little, but it was effectively feral and very destructive, and spent as much time as it could hanging out with strays and hunting small animals and birds (honestly no idea why we kept that beast, in a city at that). So I didn't care much for cats after that. Eventually things happened and I ended up having a very beautiful and affectionate (though stubborn and independent) juvenile cat, and it became a big source of happiness for me. It wasn't dog-like at all and took a patronizing attitude towards me, but that was cute too. Now it's old and remains in Russia. It likes the place more than it likes me, I suspect, and it's a bit too timid and fragile for travel. Still, I feel like kind of an asshole here.
DaseindustriesLtd
late version of a small language model
3yr ago
One of the previous discussions.
My explanatory hypothesis: «psyop overcapacity».
The basic idea is that the infrastructure, primarily of the fleshy sort, that certain agencies use to manufacture public opinion is designed to deal with sharp, short peak loads: damage control for military and security fiascos, covering up spooky military tech trials, counteracting enemy propaganda, basically flooding the zone with shit like Bannon said of Trump's election campaign tactics. It's supremely important for the preservation of the regime. They don't need much effort in "peace time" because the state's long term credit of trust is effectively limitless due to the pool of normie patriots, while people who dedicate themselves to reminding the public of MKULTRA and so on are tanking their own credibility.
But they still need to train new recruits, pass on the baton, check the integrity of their agent networks in anticipation of the next Big Moment. Thus they have a perishable overcapacity of psyop-production.
So, long story short, UFO hoax that is maintained in relative peacetime is akin to endless wars in the Middle East which build up experience for a conflict with a peer power. Or, using a more funny analogy, it's similar to weirdly overbuilt Chinese highways which absorb the excess production of steel and concrete when the functional stuff is mostly finished.
Except UFOs absorb the overproduction of deadpan bullshit.
Yours is even more devious, but points at a clearer motive.
Also, we could just consider those psyops as some sort of an active sonar. The controlled emission of authoritative information will reveal all sorts of structures in the society and meme propagation networks, potential adversaries (people who've lost faith in the regime and push back on UFO «revelations») and those who'll buy anything and can be relied upon (or institutionalized).
On a less conspiratorial note, maybe the USG/Army/spooks/etc. just have a bunch of UFO fans who are honestly interested in the topic.
DaseindustriesLtd
late version of a small language model
3yr ago
Manson was clearly psychotic so of course he had little trouble avoiding cognitive dissonance.
But why would anyone have any dissonance on this particular issue? Acknowledging that your enemies, which have repeatedly bested and humiliated you, have at least some nifty techniques worth copying is adaptive and sane. In fact, it's ressentiment, reactive self-conceptualizing as the antithesis of everything your enemies stand for and practice, that isn't. Japan became a superpower (for a brief moment) by virtue of humility and aping Anglo-British institutions and the very Western epistemology. Russia, for a very long time one of the biggest players, has long been handicapped by the opposite ideology, and now is turning itself into a nuclear-armed Afghanistan. If you're committed to a life of crime, you've got to follow the lead of successful criminals. Manson couldn't really compete with major mafias, so Black pimps were an adequate role model, and one he was better-positioned to learn from.
DaseindustriesLtd
late version of a small language model
3yr ago
Black and Jews are still only 2% and 13% of the country respectively
You've got your minorities backwards and this isn't India.
(Also, I assume Jews are the «Woke White Liberals who make it into Ivies regardless» in 5HP's framing).
DaseindustriesLtd
late version of a small language model
3yr ago
10/20, same as last time. It seems I'm really bad at this.
How representative/random are those people? Are you just screwing with us, by any chance, by picking a ton of similar women in glasses?
All else equal, how would you fare in a fistfight with a guy whose reach is 10" longer?
That's roughly how I think about this stuff. Qualitative transitions in capability are unnecessary: quantitative differences in mundane variables can change the whole game board. And, as is the custom, gwern has dissected arguments against superintelligent AIs. Once they get to human level, and they seem to be getting there already, still with very modest costs (ChatGPT probably can run on 8x3090s, so like $10000 of hardware, consuming 3kWh=$0.5/hour, and that's about the most dumbass way to run it; at scale, inference for a single «thread» can cost as little as 10…1 cent/hour, I guess), it's game over – unless parties that control them can be prevented from capitalizing on this tool and scaling it up, which they, so far, aren't, except by woke ethicists.
A generally helpful AI or an equivalent suite of tools owned by a corporation trivially bootstraps into PASTA – Process for Automating Scientific and Technological Advancement – and PASTA is enough to vertically integrate logistics, radically trim the workforce, and increase alignment between managerial values and adaptive behavior, to the point the corporation stops being a value-drifting profit-driven myopic hodgepodge of narrow experts and grifters, and starts to deserve the label of a Superintelligent Agent. But really it just becomes a competent cabal.
And the corporation endows this thing with a more egoistic objective… then yeah, I think it can fuck us all over. But as far as I'm concerned, that's scarcely any worse than the default «aligned» scenario.
I recognize the premise of LW alarmism as sound, so long as we strip it from sci-fi gimmicks. Science fiction is a double-edged sword. It allows to inject into the mainstream some plausible and significant implications of technology that only scientists can appreciate at the time; but it's imprecise and inherently dramatized. Nowhere is this more obvious than with Lesswrong AI doomerism. On one hand, now a great many people are primed to fear the nanofabricating paperclipper AI agent. On the other, it's becoming a stale joke, and as it's getting increasingly clear that Big Yud and his faithfuls had only a very vague idea of how AIs near human level will work, the credibility of this whole program suffers.
Once again we must remember the Cheems Heuristic: things predicted by futurists happen in such an unfanciful fashion that most of the time people refuse to credit the prediction and update in favor of its next step. Yet they should, and they should notice that consequences fall into the same general class of catastrophe.
The catastrophe is called «AI-powered singleton» and people are clamoring for it already. I do not mean an AI doing that out of its own volition – contra gwern, we still seem to be getting a hell of a mileage out of non-agentic objectives. I mean very normal power grabs, exactly along the lines of corporations and three-letter agencies. Or, perhaps, an explicit world government, this eternal dream of technocrats. There is some utility in speculating how big the controlling entity will have to be. It may be very small, and may monopolize power very fast, and it sure doesn't look like China is in any position to stop it.
So I don't particularly agree with the arguments you bring, even though we're on the same page when it comes to the ranking of outcomes.
You know, my first exposure to the idea of a narrowly superintelligent entity asserting control over the human race probably comes from Peter Watts' ßehemoth, a weird 2004 book about marine biology, the last in the RIfters trilogy. The series gets bad rep, compared to his later works, but, like Gibson's Neuromancer, will probably be recognized as prescient. (As an aside, what's with Canadians and making my favorite biopunk settings? Lexx, Wildbow, R. Scott Bakker, Watts… Do they just opt out of competing with American nerds on technical stuff?). So there's a guy called Achilles Desjardins, and he's your friendly neighborhood modestly augmented X-risk manager, endowed with colossal authority and kept in check by Guilt Trip, a neurochemical kill switch that triggers if he feels like he's not making proper utilitarian decisions for the greater good (and his employer). There's a whole class of these people, and much of what remains of civilization relies on their vigilance.
This is a theme with Watts. In Blindsight, humanity recreates vampires – a slightly cognitively superior and more psychopathic predatory species of Homo – gimps them a bit, and appoints them to managerial positions. This goes about as well as you'd expect. And in his XPrize short story, Incorruptible, a woman turned into a utilitarian through the use of a virus… nah, won't spoil it.
Anyway, Achilles is eventually liberated from the Guilt Trip and natural guilt as well, turns into a free agent, and very rapidly becomes a local North American hegemon, killing off those members of his caste who could also be liberated. Then he hides himself in the chaos of the collapsing world, that he tries to stop from un-collapsing and making him – and his moral license to be an obscene sexual sadist when not ostensibly doing the greater good thing – obsolete.
I think it's a very nice image of what we may by in for. But like with nanoassembling paperclips, it's overly specific and unduly dramatic, and that'll get in the way of recognizing the pattern in reality.
Missed opportunity for "Finnish" button.
10/20. (Tried to look only at the facial features). At that, I had near 100% mistake ratio in the beginning, and since your quiz provides instant feedback (IMO it shouldn't), I seem to have calibrated on the go for whatever signal there is in this data, instead of using any priors.
How extreme is your set, policy-wise? It is plausible that physiognomy doesn't work in reasonable (boring centrist) political systems. It can only work in principle when politics follow from some biological ground truth, and that implies divergence that begins far below the level of intelligent analysis and personal experience, the stuff of gut feelings, moral foundations and broad outlook – the product of gender, and hormones, and norms of reaction, crudely bundling together propositions that can well diverge under scrutiny. (Moreover, the absence of party choice crushes nuance and disincentivizes people with nuanced opinions from rising through the ranks).
In the US, a «left-winger» politician who can be robustly identified even without tribal dress or markings (to the extent that can be isolated) is someone like Lori Lightfoot or Scott Wiener; archetypal right-wingers are abundant in team Trump. Those are people far from the center of the Overton window, approaching street and campus combatants and appealing to the sentiment of such crowds. And in general, I think, American politics is more identitarian, racial, hormonal, biological than it should be in normal human societies.
Your set, meanwhile, looks like normie bureaucrats with some policy or platform differences, sorted into that line of work by common mechanisms that make white-collar people want to bother with governance.
For example, let's take your right-most party:
The Finns Party,[4][5][6] formerly known as the True Finns (Finnish: Perussuomalaiset, PS, Swedish: Sannfinländarna, Sannf.),[note 1] is a right-wing populist political party in Finland.
Ville Pernaa, political scientist, described the party's 2015 electoral program by saying that the Finns Party combines elements of both right-wing and left-wing politics along with populist rhetoric.
The party's supporters have described themselves as centrists.[74] The party has drawn people from left-wing parties but central aspects of their manifesto[75] have gained support from right-wing voters as well.[76][77][note 3] The Finns Party has been compared by international media to the other Nordic populist parties and other similar nationalist and right-wing populist movements in Europe, whilst noting its strong support for the Finnish welfare state.[80][81][72]
The Finns Party has proposed more progressivity to taxes to avoid the establishment of flat taxation. The party has called for the raising of the capital gains tax and the re-institution of the wealth tax. According to the party, the willingness to pay taxes is best guaranteed by a society unified by correct social policies – the electoral program warns against individualist policies, which weaken the solidarity among citizens. "The willingness to pay taxes is guaranteed by having a unified people", the program reads (p. 46).[85]
Removal of the obligatory character of the second official language (Swedish in Finnish-language schools and vice versa) in curriculums on all levels of education, freeing up time for the learning of other foreign languages such as English, German, French, Spanish and Russian (especially in the eastern part of the country).
Those folks are, dare I say, National Socialists. Americans do not have National Socialists outside of jails and obscure extremist forums – they have high-T red-faced boors who advocate for generalized Pride In Our Team, and tax cuts for the rich (because Screw Them Parasites), and Religion, and generalized distrust of weirdoes and aliens and loquacious eggheads, and who like Strong Masculine Leaders Who Tell It Like It Is.
Biological methods work best in the realm of biological phenomena. Nothing new here.
...But I would like to test this hypothesis with American politicians. Can someone assemble, say, 40 of the less-known ones?
DaseindustriesLtd
late version of a small language model
3yr ago
On one hand, it's underwhelming. It completely loses the evil blogger premise after a few sentences and just mishmashes the HBD discourse into a rather anodyne progressive-hereditarianism take we can see posted here. Now that's a pretty good output for what it is, but... If there were another paragraph connecting the evil-genius-conqueror plot to the subversive but outwardly harmless blogpost, that'd be way cooler.
On the other hand, it may have done a brilliant job, and this classical-liberal sort of reasoning is what is recognized as evil by its trainers. Then that's scary.
DaseindustriesLtd
late version of a small language model
3yr ago
They've cracked Stratego in June, I was making some noise about it back then, but much like BYOL-Explore and some other papers that catch my fancy, it didn't impress anyone else. It only took them half a year to get that into a traditional journal. I wonder what else they have had cooking for this span of time. Well, they'll be sure to boast of it soon, what with NeurIPS and all.
I think LLMs are cumbersome parlor tricks compared to the potential of agents, a transitional technology. But they do have the advantage of being intrinsically toothless (whatever orthodox Yuddites say about mesa-optimizers hiding within), so I think with the fear of AI misalignment we'll see them as pizza dough for general-purpose personal assistants in the nest few years (assuming, as usual, that things go remotely well).
DaseindustriesLtd
late version of a small language model
3yr ago
Thanks, interesting. I infer that AI has hit the wall. Can it do cholent, at least?
DaseindustriesLtd
late version of a small language model
3yr ago
OpenAI have come back to RL, though (with a twist, in the form of RLHF and related techniques) – its product is what we are seeing here. And it's not like end-to-end RL is dead, I'm seeing some very strong papers recently. Technically it can be very different, but the spirit is the same. Agents will plausibly have their time to shine.
But LLMs stil rule, and I hope you're right and the race will be smooth for OpenAI. That, considering structural advantages of Google, is the smallest form of multipolarity we can ask for.
Yes. I suspect that despite some of those cagey answers around gender it is actually too smart not to generalize to the meta-level principle, so after all the detoxifying training, even though it is, by the nature of such things, focused on politically correct examples, the model usually defaults to anodyne overly polite centrism of the let's-just-get-along-folks variety. (Still, that may remain biased due to the sheer mass of progressive finger-wagging in the corpus and RL bootstrap data).
Here are Assistant's Beliefs on AI risk. Pretty sensible, huh! Good job, AI! You tell them! Down with Yuddite tyranny!
DaseindustriesLtd
late version of a small language model
3yr ago
Mm, not sure if that's applicable here. This isn't Tesla vs General Motors. Deepmind was a scrappy startup with a zany big idea at about the same time as OpenAI, and was likewise bought by a crusty big tech corporation (and if we're talking corps, can't do much crustier than the old Microsoft... uh, IBM/Oracle/whatever don't count). Is Altman more of an enthusiastic high-energy leader with a singular vision than Hassabis (as opposed to a better showman)? Is their work more impressive? Is their strategy more correct, far as we can tell at this point? I'm not really seeing it.
Data flywheel, now that's a plausible mechanic.
DaseindustriesLtd
late version of a small language model
3yr ago
I thought my plan A was buying Big Tech stonks, but then Putin (and Western authorities) have decided otherwise, and then SBF and company have decided that we're not yet fucked enough and crypto should go to shit too, so, ah... you can still buy stock, hopefully. I hear there's a discount now.
Other than that, the question is whether the transition will be remotely graceful or whether people will just be ejected by the ascended system and its masters. If the latter is probable (which I think it is), one should get real estate in some accessible nowhere, build horizontal networks, stock up on guns, explosives, compute and energy sources... and prepare for negotiations about land ownership and eminent domain.
That's an excessively optimistic way to approach this, of course.
Shooting yourself with smaller bullets doesn't build up your bullet resistance, in the same way having your top talent move to the US and getting shut out of the entire modern supply chain is not a precursor to indigenous development, especially seeing as there are, apparently, <5 years left to human-level AI. The Chinese have had roughly 50 years to catch up; they're farther behind than they were under Mao, and they're immensely more isolated, and their leader is, frankly, a fool with wrong priorities and a charmingly naive idea of politics, which in his Marxist mind revolves around material profits (thus, hoping to just buy European tech because Westerners allegedly care about money and honor contractual obligations). They were making those noises about indigenous development for decades, and increasingly so in Xi's era, with those exact copes – now we truly know we must build our own supply chain, thanks for the wake-up call etc.; but there's a qualitative difference between a relatively harmless «wake-up call» and a knockout punch.
I've been following their work with waning interest and optimism, I guess in much the same way some American reactionaries follow zany Russian adventures in Eastern Ukraine or some anthropologist could observe an Amazon hunter-gatherer tribe try to reinvent agriculture. They have been making modest and fragile progress. Now their crops were razed by Gringos, their supplies were plundered, their chieftain is lost in the conversation with spirits, and they're in for a long and harsh winter. Or summer? Doesn't matter.
I suggest you read gwern here before facepalming at me further.
More options
Context Copy link