And by an LLM you mean the people who train and tune the LLM to generate the type of responses they like.

I doubt Trump will do anything that could be perceived as encouraging more whistleblowers, at least while he’s in office.

Best case—a Democrat wins the next election, Trump gives up on trying to stop it, and pardons Snowden on his way out.

If it’s in a Greek or ancient Latin context I pronounce it with a hard C, but if it’s a general English context I pronounce it with a soft C.

I’m not sure what the third way would be.

As others are pointing out, there are mass protests going on—but I think there’s more to it than that.

The general message of all protests is “listen to us or else”. In the US for the last fifty years, “or else” has been understood to mean “or else you’ll lose the next election”—but it’s becoming clear that this threat has no leverage with Trump, either because he’s confident he can manipulate elections (through whatever means) or because he intends to accomplish his goals in his current term and doesn’t care what happens after that.

So protests need to find some other goal and some other message. Right now they’re looking for other weak points (e.g., Tesla dealerships), but once it’s clear they’ve got a strategy Trump is actually afraid of, the numbers will grow.

Yeah, jpeg converts to lab (or something similar, I think). But the dimensions are the same: one channel for lightness, and then a number of channels one less than the total number of sampled frequencies to capture the rest of the color space.

Spectral JPEG XL utilizes a technique used with human-visible images, a math trick called a discrete cosine transform (DCT), to make these massive files smaller […] it then applies a weighting step, dividing higher-frequency spectral coefficients by the overall brightness (the DC component), allowing less important data to be compressed more aggressively.

This all sounds like standard jpeg compression. Is it just jpeg with extra channels?

So they’re adding phone capabilities to Tamagotchi?

The same reason we capitalize peoples’ names like that, since a title is the proper name of a written work.

Some animals are more equal than others!

Ok—to the extent that SVG is HTML, the variant of HTML that it is is a flavor of XML.

More precisely, both are flavors of XML.

Also, be sure to fully specify the location—one time I just put “Athens” and ended up in Athens, Georgia.

I like Molly White’s recent take, that it might be more productive to treat this as a labor issue instead of a copyright issue (at least in principle). Even if the AI corporations aren’t technically re-selling copyrighted works, they’re still profiting from the authors’ unpaid labor.

If there are a bunch of posts on a particular topic, shouldn’t it keep at least one of them? Otherwise it would tend to completely filter out the most significant or interesting topics.

I had a really good pizza topped with stinging nettle once.

I always insist on colors within the visible spectrum.

(4) you pass, barely

That implies that 1, 2, and 3 are all failing (perhaps with different degrees of embarrassment). If all failure is equivalent in practice, you might as well maximize the non-failing outcomes and go with A.

I don’t think we’re usually conscious of the availability of multiple narratives to the point that we can mentally simulate each one and compare their potential utility.

The choices we consciously make are the ones that arise inside of narratives, not between narratives themselves.

we will choose the one in which we profit the most

I think we’ll choose the narrative most consistent with our existing worldview, even if it hurts our own interests.

Causality, like entropy, is an emergent property that’s hard to pin down formally, but is a critical element of any narrative. I’m with you that far, but how do price tags and double breakage figure in?