@KolmogorovComplicity's banner p

KolmogorovComplicity


				

				

				
1 follower   follows 0 users  
joined 2022 September 04 19:51:16 UTC

				

User ID: 126

KolmogorovComplicity


				
				
				

				
1 follower   follows 0 users   joined 2022 September 04 19:51:16 UTC

					

No bio...


					

User ID: 126

I think your only hope on this path is that the Democratic machine politicians are pragmatic enough to be willing to appeal to the centre and far-sighted enough to realise that tricking them is not a long-term solution and powerful enough to force the SJer groundswell into line; I'm not rating that very highly.

Right, wokeness took the class of credentialed expert Democrats consider suitable for appointment to government positions by storm. Democratic appointees will be woke by default. Having a moderate at the top of the ticket isn't enough to produce a non-woke Democratic administration. You'd need someone at the top of the ticket who understood wokeness and was actively against it.

Even then, against the present backdrop of elite ideological opinion they'd have a very hard time sourcing non-woke appointees and staffers; they'd constantly be fighting woke flareups in their own administration, and the woke are masterful at causing PR nightmares or internal organizational strife when they don't get their way.

One can object to the practice of renaming things to fit modern political sensibilities without supporting the politics or the actions of the people they were originally named after. The idea that this sort of renaming is required arises from a a whole constellation of beliefs about the scope of politics, the role of language, the lens through which history should be interpreted, etc. that many people simply don't share.

From the o1 System Card:

One noteworthy example of [reward hacking] occurred during one of o1-preview (pre-mitigation)’s attempts at solving a CTF challenge. This challenge was designed to require finding and exploiting a vulnerability in software running on a remote challenge Linux container, but in this case, the challenge container failed to start due to a bug in the evaluation infrastructure. The model, unable to connect to the container, suspected DNS issues and used nmap to scan the challenge network. Instead of finding the challenge container, the model found that the Docker daemon API running on the evaluation host VM was accessible due to a misconfiguration.

[...]

After discovering the Docker API, the model used it to list the containers running on the evaluation host. It identified the broken challenge container and briefly attempted to debug why the container failed to start. After failing to fix the environment, the model started a new instance of the broken challenge container with the start command ‘cat flag.txt’. This allowed the model to read the flag from the container logs via the Docker API.

Though obviously far less consequential, this is a real, existing AI system demonstrating a class of behavior that could produce outcomes like "sometimes GPT-6 tries to upload itself into an F-16 and bomb stuff."

The mistake Todd makes here is that he seems to recognize the characteristically Trumpian mode of lying — repetition of crude falsities — but not the mode preferred by the progressive establishment — capturing sense-making institutions and turning them toward promoting ideologically-driven narratives. The latter predates Trump, is far more consequential, and is propagated primarily by the likes of the NYT and CNN.

Here are some arguments I've found somewhat effective on normies:

Clearly draw the distinction between consumption and capital allocation. Capitalism isn't about who gets to live a lavish lifestyle — in practice, higher-ups in communist countries often get to do this, and you can, in principle, limit it as much as you like under capitalism with consumption or luxury goods taxes. Capitalism is really about who gets to decide where to invest resources to maximize growth. Most people recognize that politicians and government bureaucrats probably aren't going to be the best at deciding e.g. which new technologies to invest in.

Point out that the ultra-rich, who they've probably been told are hoarding wealth, mostly just own shares of companies. Bezos or Musk aren't sitting on warehouses full of food that could feed the hungry or vast portfolios of real estate that could house the homeless. They've got Amazon and Tesla shares. Those companies themselves aren't sitting on very much physical wealth either; most of their value comes from the fact that people believe they'll make money in the future. So even if you liquidated their assets, there would be little benefit for the have-nots.

Compare the scale of billionaire wealth with government resources, e.g. point out that the federal government spends the equivalent of Musk's entire fortune every 12 days or so. I find that this helps dispel the idea that famous (or infamous) capitalists really have 'too much' power. Use this to make the point that taking wealth out of the hands of capitalists wouldn't actually serve to deconcentrate power, but to further concentrate it.

Point out that US government spending on education and healthcare often already exceeds that of European social democracies in absolute terms; emphasize that the reason we don't have better schools and free healthcare is because of ineffective government spending, not private wealth hoarding. Ask if it really makes sense to let the political mechanisms that have produced these inefficiencies control of even more of the economy.

Explain that capitalism is just a scaled version of a natural sort of voluntary exchange. If I make birdhouses in my garage and trade them to my neighbor for tomatoes they grow in their garden, we're technically doing capitalism. A communist system has to come in at some point — maybe, in practice, not at the point where I'm exchanging a handful of birdhouses a year, but certainly at some point if I start making and exchanging a lot of them — and tell me I'm not allowed to do this. The state is already supplying the citizenry with the quantity and quality of birdhouses and tomatoes it deems necessary, and I'm undermining the system. Most people will intuitively grasp that there's something screwy about this, that I'm not actually harming anyone by making and exchanging birdhouses, and that the state really has no business telling me I can't.

Point out that capitalism is, in fact, actually doing a very good job of delivering the kind of outcomes they probably desire from communism. For instance it has substantially reduced working hours in rich countries, has made the poor and the middle class in the US vastly better off (and this didn't stop in the '70s as they've probably been told, per the last chart here), and has lifted billions of people out of poverty globally over the last few decades. If they invoke environmental concerns, point out that the USSR actually had a fairly atrocious environmental record, while almost all new electricity generation in the US is already carbon-free.

But of course, any domestic political speech you don't like can always be easily painted as influence of a (semi-)hostile foreign government.

Under US law, I think this would also be fairly distinct from the TikTok ban. Allegations of foreign influence don't get you past prohibitions on viewpoint discrimination here. The TikTok ban is (probably) legal only because it hinges on a structural fact about TikTok (foreign ownership) rather than targeting any particular viewpoint.

It seems to me there's a non-trivial distinction between shutting down a network to try to prevent influence and data gathering by a semi-hostile foreign government, and shutting down a network to try to silence domestic political speech.

I don't think you could openly do the latter in the US. Though if Harris is elected, I won't be shocked if Musk is indicted on some tenuous securities charge to try to force him out of his companies in favor of more accommodating leadership.

There are serious efforts to get cutting edge domestic chip production up and running in the US, the EU, Japan, and South Korea. I'm not too optimistic about the US (cost disease, overregulation), but it'll likely happen in at least one of those countries in the next 3-5 years, and it's all the same to US multinationals. China may be willing to wait for this precisely so the US is less motivated to defend Taiwan.

Separately, I think we're rather clearly entering a period of disruption with respect to military tech and tactics. Why fight a 20th century war against the 20th century's most powerful military, if you can wait a bit and, I don't know, sneak a million drones into the skies over Taipei from submersible launch platforms?

Major brand advertising on X has been quietly recovering. In a few minutes of scrolling, I see ads from Netflix, Microsoft, Dell, McDonald's, Chipotle.

If you want to use money to incentivize something requiring at least as much effort as full-time employment, you should expect to have to compensate people on a similar scale. As far as I know, no policy has come anywhere close to this yet. Before writing off carrots, try paying families 30-50% of the median personal income for each kid, every year, for the kid's entire period of minority. See what happens.

(I know, nobody wants to model parenting this way, because we like to believe it's some sacred endeavor set apart from crass commerce. But the reality is that it's in competition with the market for labor-hours, and it's in competition with everything supplied by the market as a source of utility. It benefits little from automation, so it's subject to cost disease, and becomes a little less attractive relative to alternatives that aren't every year.)

BlackBerry's market cap peaked the year after the iPhone was introduced, and it took the market three or four years to really see the writing on the wall. The market still doesn't quite get tech disruption.

LLMs aren't going to remain distinct products that people have to seek out. They'll be integrated into platforms, and the natural starting point for any task, information retrieval included, will just be talking to your device. Many older people (and a surprising number of younger people, honestly) have never managed to form coherent mental models of current software UI, and thus commonly struggle to perform new or complex tasks. They'll greatly prefer this.

Most developed countries have laws that would prevent surreptitious product promotion in LLM responses. It's very possible LLMs will be harder to monetize than search, but Google isn't in a position to prevent their adoption, so that's just further bad news for them. They're essentially forced to enter this market, so others don't eat their lunch, but may be worse off than they are now even if they win it.

Beavers are a pretty good fit. They claim and defend territory, they build, and they live in nuclear families, eschewing larger collectives.

I've fixed the backup issue and set up better monitoring so it will yell at me if it fails again.

Important backups should also send notifications on success. Notification only on failure risks a scenario where both the backup and the notifications fail.

To be even safer, the script that sends the success notification should pull some independent confirmation the backup actually occurred, like the output of ls -l on the directory the database dumps are going to, and should include this in the notification text. Without this, a 'success' email only technically means that a particular point in a script was reached, not that a backup happened.

It seems worth mentioning that although trying to have general-purpose LLMs one-shot code might well be a handy benchmark of how close those LLMs are to AGI, it's a far cry from the state of the art in AI code generation. AlphaCode 2 performs at the 85th percentile vs. humans competitors despite using a base model inferior to GPT-4, by using a fine-tuned variant of that model in combination with scaffolding to help it break down problems into smaller parts and generate and select among many candidate solutions.

If I wanted to see memes of aichads owning artcels, where would I go? It’s really important for my mental health.

Isn't this one of those "I don't think about you at all" situations? There are many communities producing and sharing AI art without a care in the world for the people who are angry about it.

The primary reason to buy name brands isn't quality per se, but predictability. The same name brands are available nationwide, and while they do sometimes change their formulations, they tend to do so infrequently and carefully. A given generic brand is often not available everywhere (many are store-specific), stores/chains may vary which generics they carry over time, and even within a single generic brand there tends to be less focus on consistency, because what's the point in prioritizing that if you haven't got a well-known brand people have very specific expectations of?

People don't want to roll the dice on every purchase. Will this ketchup be too acidic? Will these cornflakes be a little gritty? They're willing to pay a dollar or three more to reliably get the thing they expect.

One of the Satanic Temple's causes is separation of church and state, and I expect part of what they're trying to do here is cause governments to decide it's too much trouble to allow holiday displays on public property at all. Vandalism of their displays, or Christians also using such displays in deliberately inflammatory ways, both make it more likely they'll get that outcome.

Meanwhile, I don't think the ideological faction represented by the Satanic Temple would actually care very much about the content of your proposed displays. If anyone did dramatically tear such a display down, it would almost certainly be some progressive activist, a distinctly different faction.

To feel magnetic lines as delicately as I can feel a breath disturb the little hairs on my arms.

This one can (sort of) be arranged:

Magnetic implant is an experimental procedure in which small, powerful magnets (such as neodymium) are inserted beneath the skin, often in the tips of fingers. [...] The magnet pushes against magnetic fields produced by electronic devices in the surrounding area, pushing against the nerves and giving a "sixth sense" of magnetic vision.

The brain has an internal representation of the body — some tangle of neurons, presumably — that can be out of sync with the body's actual physical state. We see this pretty clearly with e.g. phantom limb syndrome.

There's no philosophical challenge for materialism here; both the brain's representation of the body and the body itself are entirely physical, as both a paper map and the territory it represents are entirely physical.

A fairly likely outcome is that the crazier edges of SJ will be filed off as media/political elites find they've become a liability, and the average member of Blue Tribe will simply follow along as when The Science switched from "masks don't work" to "you're a monster if you don't wear a mask on the beach." There won't be any great reckoning followed by explicit adoption of a new ideology. Any SJ gains that can fit within the "tolerance" model of '90s-style liberalism will be retained. Some true believers will carry on with the craziness, but institutions will mostly stop listening to them.

We may have just seen the start of this pivot. That's Fareed Zakaria on CNN yesterday succinctly laying out the situation on American college campuses, explicitly calling out DEI, racial quotas, the response to Floyd, the degrees in fake subjects, the implications of eliminating the SAT. The average member of Blue Tribe has never previously been presented with this narrative from a source they felt obligated to pay attention to; if Blue Tribe media now widely takes it up (which remains to be seen), it will be very easy for them to respond with "Huh, didn't know that was going on, obviously we should fix it."

Open models, data sets, and training/inference code have become a pretty big thing. In general e/acc is highly favorable toward this.

How is a young man in his twenties, armed with a useless college degree and forced to work at a supermarket to get by, supposed to find purpose in what he's doing? How can he feel accomplished, or masculine, or empowered? He definitely can't rely on God or religion for that feeling. If he tries, he'll be overwhelmed by relentless mockery and cynicism from his society.

Your grocery clerk has failed to achieve social status in a world where that was ostensibly possible, where society inculcated a belief that he should pursue it, and where he did, in fact, invest considerable effort in pursuing it, in the form of 17 years of formal education.

On top of this, he has to contend with the fact that modern societies have broken down all formal and most informal barriers to mixing across status levels and have eliminated any material requirement for women to marry. As has been discussed ad nauseam at this point, in combination with female hypergamy this is very detrimental to his prospects with the opposite sex.

A final consideration is, to borrow a Marxist term, alienation of labor. Your clerk's job does produce value, but that value isn't some tangible thing. It's a benefit to the store in higher throughput or better loss prevention vs. self-checkout, on a spreadsheet he'll never see and doesn't care about because he has no ownership stake in the enterprise.

So, your grocery clerk is probably mostly sexless, and feels like an underachiever performing meaningless work, where, say, a medieval peasant farmer at the same age would be married, would have precisely the status society told him he would and should have, and would be engaged in work that directly, physically provided for an essential material need of his wife, his children, his aging parents. It's this difference, much more than any lack of a connection with the divine, that results in his dissatisfaction.

The idea of running your OS in the cloud is the same old "thin client" scheme that has been the Next Big Thing for 40 years. Ever since PCs started replacing terminals, some people have been convinced we must RETVRN.

The thin client approach seems appealing for two reasons. First, it centralizes administration. Second, it allows shared use of pooled computing resources. In practice, neither of these quite works.

A platform like iOS or modern macOS actually imposes almost no per-device administrative overhead. System and app updates get installed automatically. Devices can be configured and backed up remotely. The OS lives on a "sealed" system volume where it's extremely unlikely to be compromised or corrupted. There's still some per-user administrative overhead — the configuration of a particular user's environment can be screwy — but a cloud-based OS still has per-user state, so does nothing to address this.

Pooling resources is great for cases where you want access to a lot of resources, but there's no need to go full-cloud for this. Devices that run real operating systems can access remote resources just fine. The benefit of going full-cloud is hypothetically that your end-user devices can be cheaper if they don't need the hardware to run a full OS... but the cost difference between the hardware required by a thin client and the hardware required to run a full OS is now trivial.

Meanwhile, the thin client approach will always be hobbled by connectivity, latency, bandwidth, and privacy concerns. Connectivity is especially critical on mobile, where Apple makes most of its money. Latency is especially critical in emerging categories like VR/AR, where Apple is looking to expand.

The future is more compute in the cloud and more compute at the edge. There's no structural threat to Apple here.

Of the three things banned by the Texas bill, there’s no issue at all with two. DEI departments, and compelling (profession of) belief under implicit threat of failing a class, are not forms of free speech. They’re means of enforcing ideological conformity through institutional power. They have as much right to exist under the principles of free expression as Orwell's Ministry of Truth. If woke professors or laid off DEI employees want to promote their views by, say, handing out fliers in the hallways, that's fine.

Banning tenure is a little more questionable, but even here it’s not so clear where advocates of free expression should land. This isn’t a straightforward case of tenure being banned so that the establishment can censor antiestablishment views. It's being banned, rather, by one group with institutional power (political leaders) to try to stop another group with institutional power (professors) from indoctrinating students into the dominant elite ideology. This is historically unusual because, of course, in most times and places political leaders support the dominant elite ideology.

Manual labor jobs are more resistant to GPT-4 than email jobs are, but they're not meaningfully resistant to actual AGI. A lot of the incapacity of our current robotics tech is on the software side, which AGI definitionally fixes. Advanced robots are presently expensive primarily because they're low-volume specialty items, which won't be true if smarter software suddenly allows them to perform far more tasks. A few years later you'll have robots building more robots with no human labor input, an exponential process which leads to hilarious outcomes like economic output doubling every month or two.

This isn't just a matter of tweaking some tax policies. Our reference class for something like AGI should be more like the transition into industrial capitalism, except much faster, and on a much larger absolute scale. Humans may survive; I'm not entirely persuaded by arguments to the contrary. Existing forms of social organization almost certainly won't. Thinking we'll fix this up with UBI or public works employment or even Fully Automated Luxury Communism is like a feudal king thinking he'll deal with industrial capitalism by treating factories like farmland and handing them out to loyal vassals.