oceanofsolaris

I don’t think you can really accuse the lesswrong crowd of being in it to hype up OpenAI.

They don’t profit from it (well at least the current doom crowd), they have been saying that this is super-risky since “forever” (15 years), they strongly argue against e.g. OpenAI going private … I really don’t understand where this whole strand of thinking of hidden ulteriour motives for AI doomers comes from, AFAIK there was never a single clear case where this happened (and the arc of people invested in AI is they they become less of a doomer the more financially entangled they become with AI, see Elon or Altman).

I am not saying that you shouldn’t build trust with your kids (it’s important and only fair to them … they rely on and usually adore you, you should in turn be kind and reliable to them and love them back).

But: the null hypothesis for anything which correlates childhood behavior, childhood environment and adult behavior should be that everything is 100% genetic. From my glance at the studies, the marshmallow experiment was instead always done with the null hypothesis that kids and their parents are only connected by living in a common household (aka “shared environment” in the sense that it’s shared between siblings).

I would still imagine that some effect of the household on the “delayed gratification” would persist, but it would probably be weaker and I am not sure it would predict much about outcomes as an adult.

Also interesting: Kreisky totally didn’t resign after the vote ended against Zwentendorf…so the whole thing was for nothing in the end.

The operator of the plant assumed that the politicians would get to their senses at some point and kept it in operational shape for some years even though it wasn’t producing any power. Then they went bankrupt because pretend-running a nuclear plant is expensive.

Can‘t you do it the old-fashioned way and call the phone? If it’s on silent, it might make it harder, but it still makes a buzzing sound since it vibrates.

Basic health insurance in Switzerland is “private” but very regulated:

All companies offer completely exchangeable plans and are only allowed to take age, municipality and desired copay into account when calculating your premium. Every resident of Switzerland is also obligated to have this basic health insurance (if you don’t, the government just picks a random insurance for you and signs you up retroactively).

I would say it’s a great system…but Swiss health care costs are also among the highest in the world (and rising steeply), so it’s not a silver bullet.

To be fair, this particular system is probably not to blame for the high costs, but it also didn’t prevent them (though it leads to a bit of innovation around e.g. telemedicine, which is great).

If a lack of capital investment (or skilled companies) is the problem, that’s a great solution.

But that’s not the reason for expensive rents in big cities.

It’s because many (rich) people want to live there … and due to physical and regulatory limits you can’t build more housing, even for piles of money it would be worth.

I don’t totally get this point. The main problem in big cities is the quantity of housing as well as the distribution of it.

For a train system, a non-state actor will have a hard time getting one of the ground (huge upfront costs, will often make a loss, cooperation of politics super important) and if they do … they will have a monopoly (which also sucks). So having transit publicly owned in some way makes a lot of sense.

But for housing, none of that is true. It’s being built (the amount built mostly depends on government approval, there is no shortage of capital investment), there is no monopoly… what is the problem that more public housing would solve?

The (admittedly outrageous) rent prices in Zürich come from a mismatch of demand and supply. And since Zürich has a lot of rich inhabitants, enough of them are willing to bid up the market rent prices to the current level.

Having a non-market solution for this might lead to cheaper rents, but introduces a lot of other losses and crucially also strangles the supply even more (look at what happened in Berlin when they introduced their moronic rent control law).

But SF is a damn rich city (due to all the tech companies and their employees). Why didn’t they solve or at least tackle these problems?

It’s not like Cruise / Waymo could change city planning against the cities will, even if they wanted.

It is true that heating should be electrified … but it’s important to note that this makes the Dunkelflaute problem more serious, not less.

Because now your inflexible base-load has increased. And while spreading generation over a wider area helps, it rapidly reaches limits (unless your power grid extends close to the equator).

Also „Fiasco“, which is depressing and great.

For a more recent take on a similar theme, see also „Blindsight“ by Peter Watts.

I assume dysgenics (which you’d assume is happening if family size is inversely correlated with education and or income).

AFAIK there are some studies that show that dysgenics are at least not the only effect in play (e.g. this study https://www.pnas.org/doi/full/10.1073/pnas.1718793115). But I am far from an expert on the topic.

And if they are wrong (and this sells well) … they might just go bancrupt, leaving you with only a small part of the upside.

What do you hedge your 100 year financial vehicle with? AFAIK there just isn’t anything that allows you to hedge against interest rates 50 years in the future.

So while this probably works out (especially if there is high inflation or high interest rates for some of the 100 years), I think the upside as a borrower is very limited, especially if this ends up being popular (if interest rates fall to 1% next year and stay there for 100 years, you didn’t get a great deal…you most likely just have a CD in a bancrupt institution).

But that’s not the default behaviour of these models at all. Instead you have to work pretty hard to extract these original images.

When using the models normally, they do generate new content, even if the style and subjects are of course interpolated from training data.

The “3d object that changes with a variable (like time)” metaphor helps a little bit, but doesn’t really help to e.g. understand how a rotation of a 4d object plays out.

Couldn't this also be solved with a freezer on the island (+ shipping larger quantities of these drugs regularly via whatever process usually brings parcels onto the islands).

I am not complaining about a drone highway, sounds great. It also sounds like quite an oversized hammer for that specific problem though.

I think that it’s weird to aggregate salaries like this without taking into account that there is a huge amount of variance between companies.

For concrete companies, it’s easy to look up salaries in Glassdoor etc. and you can see that FAANG pays much better than this, even in Germany. Smaller companies on the other hand probably often pay worse than this.

But that is just an implementation detail of who is allowed to initiate outgoing transfers from your account. For me, I needed to send a signed letter to my bank to allow recurring debiting to my landlord (and since I stopped that, no entity except myself is allowed to initiate outbound transfers from my account).

That's not to say the Taler model can't work or anything, just that having one number for inbound and one for outbound transfers isn't really necessary for basic account security. The "everyone who knows my account number can draw money from my account" problem seems weird to have (and trivially solvable) .... in which country is this the default?

China needs five to six years from starting to build a nuclear reactor to making it operational. I don’t see why western countries should be inherently unable to do this. They currently don’t want to (because they don’t need to and people dislike nuclear power).

But if they cared, they should be able to.

AFAIK the running out of fuel would be a non-issue with breeder reactors, which are disliked for the proliferation risk they are. Well, trade-offs.

Also: currently, nuclear fuel cost is a small part of running a reactor. If nuclear fuel follows a similar cost/available quantity curve as other geologic resources, we should be able to find more once we start looking in earnest.

I assume because their data source is not great (aggregation of some kind of Glassdoor data).

* Looking at data for Google in San Francisco, Glassdoor says that average total compensation for a "Senior Software Engineer" is 230K $, whereas levels.fyi (which I trust a bit more) says that Level 5 (the one labeled "Senior SWE") earns 350K $ on average. Interestingly, Glassdoor also offers a breakdown of total compensation into the components. Adding the average value of the components together also results in something around 350K $ ... I guess Glassdoor just aggregates things in a wrong way. * People at different stages of their career probably have a very different likelihood to create a Glassdoor account (let alone one that accurately names their salary). Similar with small vs. large companies, European vs. American companies etc. This potentially introduces enough bias to make the whole thing worthless for jobs that have a large variance in salaries (such as "director of engineering"). * I still assume that the general thrust of the article is right (when starting your software eng career, the bay area is not a bad place to be). But I wouldn't trust the salary numbers very much.

The underlying problem these elaborate prompts seem to solve is that the internet contains many pictures, few of which look very beautiful.

If you look at all internet pictures of sheep, many of them will not be very exciting and depict a low contrast sheep in a foggy landscape.

So to get a picture with strong saturation and clear lines, it helps to put text there that is usually associated with pictures that have these ... like "HD wallpaper" or "made with unreal engine". Most "wallpapers" might be of dubious artistic quality, but muted colors and a lack of saturation will generally not be their problem.

This is of course not the only problem with the model. It doesn't even produce a clear image of a sheep .... but that will probably get better with larger models and more training. Similarly it doesn't seem to have a sense of overall composition and tends towards fractal or tiling-like images. But those problems are probably orthogonal to the fact that the model doesn't per se try to make good pictures ... just average ones for the description you give it.

I am especially impressed by how much more realistic the lighting ends up looking despite the network probably having less information to work with than the engine! Though it ends up turning the lighting very flat. I guess that is an effect of the source images they were using.

As others have said, using this directly in the engine (tuned to work with the intended art style etc) could probably produce almost miraculous results, if it can be made to work at a reasonable frame rate.

That would also allow the developers to use high-quality rendered images instead of these green-tinted, low-contrast "automotive grade" camera images as source of truth (there are good reasons these images look like that ... but they don't look pleasant).

I would imagine that the amount of available calories is overall the larger influence on which society grows and which doesn't. Remember that in most hunter-gatherer and farming societies, population stayed constant despite high birth numbers.

AFAIK there isn't much genetic engineering involved in that project at the moment. They just rear male mosquitos that are infected with a Wolbachia bacterium (making them sterile).

I think there are still some unknowns (number of infections without any symptoms, etc.), but so far, this approach doesn't seem viable: If we keep infections low enough that the health care system can deal with the severe cases, it will take us more than two years to develop herd immunity. By that time, we will (hopefully) have a vaccine anyways.

So the better strategy might be to not drive the health system to its brink (also not clear how well it could sustain that strain for long anyways) and instead keep infection numbers very low until we have a vaccine (while opening society and the economy as much as possible).

In each case, we need to keep R0 more or less exactly at 1.0 . If it is any bigger, we have exponential spread which will quickly overwhelm the health care system. If it is much smaller, we have gone overboard with containment measures and should relax them to ease the burden for society. The only difference is how much strain is put on the health care system (which is usually already operating close enough to capacity without a pandemic).

I am not a huge fan of this (or the WaPo) simulator, since I they seem to have chosen their models for the nice look (regular grid, bouncing balls) instead of for their accuracy.

While you can give people a rough idea of how different containment efforts will work, the models are so far removed from reality that I don't think that it really helps very much. Especially the WaPos'comparison' between different containment measures is IMHO borderline negligent without putting huge caveats in front of them.

Since he seems to be married (according to Wikipedia), this strikes me as a very uncharitable reading. (It would also be uncharitable if he wasn't married though)

They mention a couple of features that are not just "long exposure" in the post:

* Compensating for moving stars

* "Live viewfinder" during exposure

* Selectively darkening the sky

* Dark current compensation (though that is probably needed for all long-exposure photography...still, not a simple "more exposure" feature)

I think this refers to Nowak et al and I assume refers to the discussion mentioned in the Wikipedia article of the book (https://en.wikipedia.org/wiki/The_Selfish_Gene#Units_of_sele...).