I liken it to a professional basketball player with a low free throw percentage. If they’re still on the team and in the league despite missing 3 free throws a game, they must be really good at the other stuff.

I would think that accurate color representation would’ve generally required the bright lights and broad spectrum coverage of sunlight, so I imagine people just…painted during the day, by daylight.

There are lots of benefits to using base 12 for measurements.

12 is better than 10, I’ll give you that. But 100 is better than 144, and 1000 is way better than 1728.

And that doesn’t even get to 0.1 versus 1/12, or 0.01 versus 1/144.

So 12 might be a better standalone number, but it’s a terrible base to work in.

Fahrenheit today is literally defined through Celsius

The same as pretty much every unit they use

At this point, that’s basically every unit other than the seven fundamental units. Degrees Celsius is defined from the fundamental unit Kelvin.

Plus the actual definitions of those fundamental units were defined based on historical measurements tied to former definitions. Today the second is defined around the frequency of the cesium-133 atom, but it was traditionally measured as 1/(60 x 60 x 24) of the time of a single rotation of the earth, which stopped serving us when we realized the rotations had too much variation between days. The meter is currently defined around the speed of light and the second, but was previously defined in terms of what they thought the Earth’s circumference was, and then a metal bar they kept in Paris, then based on the wavelength of light emitted from a transition in krypton-86. Same with the kilogram, currently kept at Planck’s constant but previously based on a particular chunk of metal that was mysteriously losing mass over time, and before that defined from the density of 4°C water and the definition of the meter.

Conventions are important. The history of how we got to particular conventions can often be messy.

Or, if the app has the private key for decryption for the user to be able to see the messages, what’s stopping the app from copying that decrypted text somewhere else?

The thread model isn’t usually key management, it’s more about the insecure treatment of the decrypted message after decryption.

JPEG Organisation?

The G in JPEG already stands for “Group.”

Ok, first off, a lion…swimming in the ocean? Lions don’t even like water. If you placed it near a river, or some sort of fresh water source, that’d make sense. But you find yourself in the ocean, a 20 ft wave, I’m assuming it’s off the coast of South Africa, coming up against a full grown, 800 lb tuna with his 20 or 30 friends. You lose that battle. You lose that battle nine times out of ten. And guess what, you wandered into our school of tuna and we now have a taste of blood! We’ve talked to ourselves. We’ve communicated and said, “you know what? Lion tastes good. Let’s go get some more lion.” We’ve developed a system, to establish a beachhead and aggressively hunt you and your family. And we will corner your, your pride, your children, your offspring…”

How ya gonna do that?

We will construct a series of breathing apparatus with kelp. We will be able to trap certain amounts of oxygen. It’s not going to be days at a time, an hour, hour 45. No problem. That will give us enough time to figure out where you live, go back to the sea, get more oxygen and then stalk you. You just lost at your own game. You are out gunned and outmanned.

The only solution is to train LLMs on more 4chan content, surely there will be no side effects from that.

Using space elevator technology (metal structural beams and metal guy cables) I think we can get things up to 100m geosynchronous “orbit” pretty easily.

I’m giving some reasons why turning on or off location services at the OS level doesn’t appreciably change battery life.

You can turn off higher level location services at the OS level, but at the radio level the cellular network will always need a precise enough location to handle tower handoffs and timing issues between the tower and phone, as well as modern beam forming techniques where the tower “aims” the signal at the phone. The simple act of the phone communicating with a specific tower tells the phone where it is (sometimes with surprisingly high precision).

911/emergency services also use more low level location techniques, but I’m pretty sure those functions don’t get called unless you dial an emergency number.

It’s not feasible for a mass market consumer product like Starlink.

Why not? That’s a service designed to serve millions of simultaneous users from nearly 10,000 satellites. These systems have to be designed to be at least somewhat resistant to unintentional interference, which means it is usually quite resistant to intentional jamming.

Any modern RF protocol is going to use multiple frequencies, timing slots, and physical locations in three dimensional space.

And so the reports out of Iran is that Starlink service is degraded in places but not fully blocked. It’s a cat and mouse game out there.

I’d think that there are practical limits to jamming. After all, jamming doesn’t just make radio impossible, it just makes the transmitter and receiver need to get closer together (so that their signal strength in that shorter distance is strong enough to overcome the jamming from further away). Most receivers filter out the frequencies they’re not looking for, so any jammer will need to actually be hitting that receiver with that specific frequency. And many modern antenna arrays rely on beamforming techniques less susceptible to unintentional interference or intentional jamming that is coming from a different direction than where it’s looking. Even less modern antennas can be heavily directional based on the physical design.

If you’re trying to jam a city block, with a 100m radius, of any and all frequencies that radios use, that’s gonna take some serious power. Which will require cooling equipment if you want to keep it on continuously.

If you’re trying to jam an entire city, though, that just might not be practical to hit literally every frequency that a satellite might be using.

I don’t know enough about the actual power and equipment requirements, but it seems like blocking satellite communications between satellites you don’t control and transceivers scattered throughout a large territory is more difficult than you’re making it sound.

90GB of both RAM+NAND combined. I’m guessing most of it is actual persistent storage for all the stuff the infotainment system uses (including imagery and offline map data for GPS, which is probably a big one), rather than actual memory in the sense of desktop computing.

Everything else that you said seems to fit the general thesis that they’re making a lot more money selling to AI companies.

If those reasons were still true but the memory companies stood to not make as much money on those deals, I guarantee the memory manufacturers wouldn’t have taken the deal. They only care about money, and the other reasons you list are just the mechanisms for making more money.

It’s a very common complaint among people administering websites. This particular AI poisoning service seems to be directed at those people.

So maybe it’s not the majority of complaints about AI, but it’s a significant portion of the complaints about AI from site administrators.

The Fediverse is designed specifically to publish its data for others to use in an open manner.

Sure, and if the AI companies want to configure their crawlers to actually use APIs and ActivityPub to efficiently scrape that data, great. Problem is that there’s been crawlers that have done things very inefficiently (whether by malice, ignorance, or misconfiguration) and scrape the HTML of sites repeatedly, driving up some hosting costs and effectively DOSing some of the sites.

If you put Honeypot URLs in the mix and keep out polite bots with robots.txt and keep out humans by hiding those links, you can serve poisoned responses only to the URLs that nobody should be visiting and not worry too much about collateral damage to legitimate visitors.

What’s crazy is that they aren’t just doing this because they make more money with AI.

No, they really are making more money by selling whole wafers rather than packaging and soldering onto DIMMs. The AI companies are throwing so much money at this that it’s just much more profitable for the memory companies to sell directly to them.

That’s why “bullshit,” as defined by Harry Frankfurt, is so useful for describing LLMs.

A lie is a false statement that the speaker knows to be false. But bullshit is a statement made by a speaker who doesn’t care if it’s true or false.

If I am reading this correctly, anyone who wants to use this service can just configure their HTTP server to act as the man in the middle of the request, so that the crawler sees your URL but is retrieving poison fountain content from the poison fountain service.

If so, that means the crawlers wouldn’t be able to filter by URL because the actual handler that responds to the HTTP request doesn’t ever see the canonical URL of the poison fountain.

In other words, the handler is “self hosted” at its own URL while the stream itself comes from the same URL that the crawler never sees.