l'empathie mécanique @dpwiz@qoto.org

@calmeilles

https://ihl-databases.icrc.org/en/customary-ihl/v1/rule28

> Rule 28. Medical units exclusively assigned to medical purposes must be respected and protected in all circumstances. They lose their protection if they are being used, outside their humanitarian function, to commit acts harmful to the enemy.

> According to the Commentary on the First Geneva Convention, examples of acts harmful to the enemy include the use of medical units to shelter able-bodied combatants, to store arms or munitions, as a military observation post or as a shield for military action.

@msinilo related skill - noticing you're confused and not letting it slide.

But maybe that's what you've described. Then the related one would be noticing being surprised and rechecking your assumptions.

@nikitonsky Rounded rectangles are the nature's purest form.

@abcdw I'm still waiting for the Codeberg to move to Guile.

@glc > I find this literature irritating and opaque.

That's a promising start! (8

@glc Perhaps. I just hope this not another "X is/has/... Y" claim.
What's your favorite or most important consequence of this distinction?

@glc @gregeganSF @acowley No, bytes/tokens/words/whatever is irrelevant. The important part that's wrong in the "word-space" model is that it misses the context. The "language" part is a red herring. What's really going on is a tangle of suspended code that's getting executed step by step. And yes there are concepts, entities, and all that stuff in there.

@futtta @QasimRashid is there an archived Hebrew original?

@glc @gregeganSF @acowley no, the LLMs aren't operating in **word**-space.

@glc @gregeganSF @acowley But that is not the word space they're walking...

@tonyg but I'm up for more interesting tests. Just tone down our claims, 'right?

I sometimes feel like the academia is ad-supported - getting distracted into putting the glaze to the detriment of the substance.

@tonyg I don't like how they buried the models they did test. A had to dig into it just to confirm my suspicion that those were indeed obsolete. Kudos for Sonnet 3.5 for standing out though.

@aras Shifting enums are the worst. But then enums with the holes in them are also bad.
C enums -- a double-edged sword with no hilt.

@ericflo I had so much hope for it. Out of the crooked timber etc etc

OTOH maybe we have been lucky by avoiding some shitty ontologies getting baked in and instead let the content speak for itself.

@nivrig Meh. No concrete predictions or causal models. Classic punditry...

@tristanC > But at the end of the day, even though it is in fact possible to order custom designs from fabs like Samsung or TSMC, unfortunately, they are still very expensive.

You'd be delighted to know that https://tinytapeout.com/ exists (=

@light Using as "updating my nodes once in a while", yes

@nikitonsky what kind of a transparency report this is without even a mention alpha channels?

@feliks games! https://haskell-game.dev/games.html