I occasionally play around with Blender a bit, so I went to see if I've got anything interesting to show, and found I rarely actually end up rendering stuff out. So then I went through my Blender files, and rendered some stuff out. So, in no particular order:

• This was a sort of "bat signal" like thing. It's projected onto volumetric clouds, which was way harder than I thought it would be:

• I honestly forgot about this one:

• This was an attempt at precisely replicating a physical object. Its a drone motor, but I didn't end up finishing all the parts:

• Another replica, this time a ham radio dummy load (that's a resistor with fairly precise impedance, used for testing). I don't have the font anymore (and can't be bothered to fix it), so the text is gone:

• Blender nearly froze trying to render this, don't know why:

• Apparently, thin film shaders don't end up working like thin film filters would in real life, and Blender gets annoyed at light sources with 1 GW power (I'm not quite sure what I was trying to do here, so I can't really say if this experiment was a success or not):

ammo is much harder to manufacture

I'm not sure if that's true. I don't have any experience with it myself, but lead is notoriously easy to cast at home, because of its low melting point.

So it seems to me, all it takes to make some (shitty) ammo is some lead, a mould (which you can make from clay) and an explosive.

For a proper modern gun, that last piece is probably pretty difficult, and you would need to make casings somehow too (I'm not sure how difficult that would be). But if you were using some form of (possibly homemade) flintlock or similar, the gunpowder from fireworks should be enough.

I think regulating both guns and ammo is the way to go.

I think there are three main issues (though I might be overlooking something):

1. AI taking over peoples work is as much an issue for mathematicians here, as it is for programers, artists, writers, etc. elsewhere.
2. We still need to check if these 700 papers are correct (AI makes mistakes at least as often as humans, usually more often). This involves a huge amount of effort. And if it turns out that they contain a bunch of mistakes, OpenAI is unlikely to care. To them, these papers have already served their purpose (generating headlines).
3. Solving a problem in mathematics means being able to (formally) convince other people that your opinion is correct (massively simplified, but imo that is what mathematics boils down to), by getting them to understand why your opinion is correct. Having a pile of linear algebra spit out 700 texts about certain math problems removes the whole "human understanding" part of that.

However, it's not entirely unprecedented for a computer to provide a proof that we don't completely understand. I forgot the details (and I'll edit them in if I can find it), but there was at least one problem that was solved using a computer and brute force trying millions of cases. That produces a proof much longer than any human could read in a lifetime, and iirc there where quite a few mathematicians unhappy about it at the time too.

This is actually something that a turing machine is suitable for.

Get youself something to remember what state you're in, write the rules on a sheet of paper and use a stack of paper as the tape, with each piece of paper being one cell.

Turing's idea was to build a mathematical model that mirrored how humans go about large, complex tasks.

Honestly, all of these super basic mathematical models for calculation are fairly well suited to being "run" by hand.

Random-access-machines can be done similarly to turing machines. Write the program down, keep the program counter on one piece of paper, and have a (numbered) stack of paper as memory.

And lambda-calculus is probably the easiest. Just write out the expression, and then write out each reduction step. This is what functional programming languages build on, and with the additional effort of checking types, you could probabl evaluate something like Haskell by hand pretty well.

All of those are turing complete (if you ignore infinite memory).

EDIT: spelling

To all the people who read neither the article, nor the other comments pointing it out:

This is NOT an AI chatbot!

And I think it's a great idea! And Kit looks cute!

Maybe Mozilla finally noticed that most Firefox users are not all that enthusiastic about AI and would rather see development on other parts of Firefox.

Nope, David Revoy does a lot of comics in this art style, and the Avian Intelligence parrot has been a recurring character in the last few.

I can also really recommend his comic series Pepper & Carrot, it has some very cool worldbuilding.

That page lists attacks in general, and not just at schools but also universities. Out of the 25 attacks listed for Germany (so since 1990), only 11 are shootings, and 14 have no deaths at all. So, you're right, Germany does know what school schootings are.

But compare this to a total of 67 (damn that number!) in the same timeframe in the UK ( https://en.wikipedia.org/... ), despite the UK containg about 20 million less people than Germany. Unfortunately, Wikipedia doesn't have a comparable list for France.

But both of those numbers just cannot compare with the US, where just the school shootings that Wikipedia lists (not all attacks, just shootings) add up to 882 over the same time frame. And its not due to higer population, the US only has about 4 times as many people as Germany.

Sorry, I realise you're kinda just trying to say, school attacks do happen in Germay too, and you're right. But you sent me down a rabbit hole. And the number of school shootings in Germany is so low that as a parent you can reasonably assume it won't happen (and they are rarely big enough to land in the news), whereas in the US it seems to be something that everyone thinks about, that schols plan for, and that has a not inconsiderable chance of happening.

Edit: spelling

German who grew up in the 21st century here. The myth that eating lots of carrots will improve your eyesight/nightvision significantly is still more or less alive and well here. I sort of believed it until I saw a variation on this meme for the first time.

I'd say it seems to have worked pretty well.

If you run your services off of a homelab rack that dims the lights whenever you power it on

If you are in this situation, then you definetly should get some more power, or at least a UPS to make sure you don't trip a breaker.

One thing I don't like about the "master key" metaphor: I do lockpicking as a hobby. And locks built for a master key are easier to pick, because you can open them with two keys. It seems kinda obvious when you think about it. (You gotta be careful when picking mastered locks though, the master wafers can fall into the keyway and permanently destroy the lock.)

Well, in English, UTC is an acronym for Coordinated Universal Time, and in French its Temps Universel Coordonné.

Now you may have noticed that neither of those actually makes UTC. This comes from a previous naming pattern for universal time (UT1, UT2, ...) (see https://en.wikipedia.org/... ).

See my other comment (sorry, Voyager seems to generate weird share links).

While violence at schools does exist in Europe, the numbers just don't compare to the US.

And don't forget! The US has stabbings too, its just that you have so many shootings that a stabbing won't make the news.

Also, what are "grooming gangs"?

amateur radio was illegal to encrypt

Amateur radio is still illegal to encrypt (with some exceptions for controlling satelites), because private communication isn't the point of amateur radio.

Besides, (in most countries) there are some topics that are illegal to talk over amateur radio about, mainly stuff like politics and religion. You're also not allowed to offer telecommunication services (i.e. pass messages on for others). Enforcing those sorts of laws would be impossible with encryption.

But to answer your question: I think we probably wouldn't have had an internet. Authoritarian regimes thrive on stability and maintaining the status quo, I think someone high up would have quickly decided that developing that sort of tech is too risky.

What exactly makes this stand out as AI to you, and not just cosplay?

I can't really find any of the telltale signs of AI. The books on the shelf all look fine, the mirror on the left edge looks perfect and I can't find anything immediately noticeable on the costume.

If this is AI, I think it's probably just the face, maybe the whole person, but certainly not the whole picture. But maybe I missed something, please let me know.

Well, if its unbreakable, then its either snake oil, or some variation on the one-time-pad (really depends on your definition of unbreakable).

But having read the article, this is definetly snake oil.

I didn't read more than what's in OP's post, but I think the reason the researchers can be so sure is because there are ways to mathematically prove that something cannot be calculated by an algorithm (this is related to how we can mathematically prove that some things cannot be proven).

One classic, simple example of this is the halting problem. It boils down to the fact that we can prove that there is no algorithm that can take any algorithm as an input and determine if that algorithm will finish (halt) after finite time.

Running Docker inside a VM is the recommended way to run Docker on Proxmox.

Personally, I just run Docker inside LXC containers. Aparently, this is supposed to be not entirely stable, but its never given me any problems. (I commit an entirely different sin against efficiency, by keeping each service in its own LXC container, each with its own Docker host.)

I also run Immich (in Docker inside a container), and haven't noticed any performance issues. But I'm the only user, and I don't use it heavily, so ymmv.

Ok, so this sent me down a brief rabbit hole on Wikipedia (https://en.wikipedia.org/wiki/Headlamp), from which I learned that the US is the only industrialzed country who have a different standard for headlights.

The US SAE standard allows for much more glare on low-beams than the ESE standard used in other countries. That explains why I see so many complaints online over low-beam headlight glare, something I have only really found to be a little annoying over here in Europe. (Also, combine the "worse" standard with significantly larger cars to get even worse glare, I guess.)

But what I was actually looking in the article (didn't know what it was called) was Glare-free high-beam, which is a system that dynamically shades and lights parts of the high-beams to provide as much light as possible without lighting up other cars.

From experience using it, I'm kinda torn. On the one hand, it does work, is really cool, and makes driving at night quite a bit easier and safer. On the other hand, it's not perfect, and sometimes (not often, but still) it takes a bit to recognize another car or fails to recognize it completely. And when that happens, I am effectively blinding that other driver with my high-beams, which is bad.

So far, especially on the Autobahn, it works well enough that I tend to use it, because being able to see nearly everything brightly at night is so much safer for me, and the few times it fails I can manually intervene.

Great write up.

As an amateur radio enthusiast, I'd like to add a bit of info to the radio related tips:

Wifi Motion & WhoFi

Depending on how it works, a radio jammer will almost definetly make you more identifiable and not less. Think of it a bit like being in a dark room. You can sorta see other people (or their silhouettes), but if someone turns on a torch, then you can definetly see the torch. (And any jammer with enough power to "blind" a wifi router at any more than a few meters away only works at 2.4 GHz, or is probably illegal.)

Radio

(or "locating recievers") This is physically impossible.

Now locate a radio transmitter is easy. Ham radio clubs often do regular "fox hunts" where teams try to find a hidden transmitter as fast as possible.

However, most digital devices are transmitting and recieving at the same time, for wifi or bluetooth. Even if you're just downloading something, you are still transmitting (I think that's probably what you meant).

Another note on Wifi and Bluetooth

This might be common knowledge here, I think I've seen a post about it before. When wifi/bluetooth are on, your phone actively searches for nearby networks/devices to connect to. It specifically looks for known networks/devices, and fingerprinting your phone based on which networks/devices it knows is incredibly easy.

So, when you aren't using it, turn wifi/bluetooth off.

Also a note on Gait Recognition

Walking funny usually doesn't work.

(sorry, no sources here, I'm in a bit of a hurry)

This looks a lot like floating point errors. There's probably some math involved in exporting that causes these slight inaccuracies.

It probably isn't too hard to write a short python script to go over the exported file, and round those values.