robryk @robryk@qoto.org
I enjoy things around information theory (and data compression), complexity theory (and cryptography), read hard scifi, currently work on weird ML (we'll see how it goes), am somewhat literal minded and have approximate knowledge of random things. I like when statements have truth values, and when things can be described simply (which is not exactly the same as shortly) and yet have interesting properties.
I live in the largest city of Switzerland (and yet have cow and sheep pastures and a swimmable lake within a few hundred meters of my place :)). I speak Polish, English, German, and can understand simple Swiss German and French.
If in doubt, please err on the side of being direct with me. I very much appreciate when people tell me that I'm being inaccurate. I think that satisfying people's curiosity is the most important thing I could be doing (and usually enjoy doing it). I am normally terse in my writing and would appreciate requests to verbosify.
I appreciate it if my grammar or style is corrected (in any of the languages I use here).
Joined Nov 2020
@rysiek I think this is similar-yet-distinct to your recent thread on various ML-based generators of something.
> Teachers will have all sorts of problems with students claiming ChatGPT-generated texts are theirs.
This might be not all bad. Too often teachers use "able to write an essay on X" as a proxy for "understands X well enough to use it". ChatGPT proves by contradiction how that doesn't follow, and will hopefully push teachers towards using better proxies (or, sadly, into using ChatGPT detectors).
@Mara Do I understand correctly that this means "you do not have to poll this future to completion, but if you do you must use the result"?
Mastodon or Fedi? I will answer about fedi.
First, it's going to go in the direction email went: we will need increasingly more anti-spam heuristics, and we'll grapple with the problem of threads not having an owner who could be doing moderation on them. We will continue to have amusing reply propagation semantics, and probably will have instances do more polling to circumvent them. People will continue to be annoyed by not having a way to subscribe to a thread other than by having something poll it, so instances will implement such polling. There's likely going to be a discussion on whether such polling is reasonable, and problems caused by traffic spikes caused by badly spread out polling.
If large instances by well-known companies will appear, fedi will be split on the question of federating with them and there'll be at least some amount of defederating-due-to-their-lack-of-defederation.
@lilithsaintcrow Do I guess right that authors are paid a fixed fraction of price on ebook sales?
@eta And what is in the place where the mask was supposed to go instead of in the fridge? :)
@ancientjames The Mastodon filter functionality reportedly admits regexes, so `.{200}` should match (part of) any post that's at least 200 characters (bytes? I hope it's characters) long.
robryk
boosted
So I've been building a 100% analog polyphonic synthesizer with an unique twist. To use only vacuum tube era technology from the 1930s.
Over 300 neon gas diodes create the sound you hear. Pretty awesome for technology from 100 years ago.
Still a work-in-progress, but I wanted to post a video of it with the innards spread out across the workbench. : }
I call it the "Neon String Machine"
I agree that having _a_ way to communicate one's language preferences via a unified in-browser UI (or browser's guesswork) would be a good thing. I don't think we can come up with something that wouldn't devolve to site-specific settings (e.g. with site-specific sets of types of content).
@ancientjames I would expect a viewer for fdm to simply ignore e.g. spindle on/off commands, and if chatgpt was borrowing from some g-code for milling/boring/... going up and down might not be that silly.
@ancientjames Maybe it was drilling a hole :P
(Were there maybe some non-FDM-appropriate codes in the result?)
In constrast to Accept, I think that Accept-Language doesn't fit reality that well.
Let's assume we are in an ideal world where users can cause their browsers to emit Accept-Language values exactly like they want (in particular different ones for different sites), and UI affordances for temporarily/permanently/... switching that for a given site/... are intuitive for all.
Even in that world Accept-Language is insufficiently expressive. Let's say we have a site with UI translated into many languages and with automated translation for contents. There's no way to express a preference between "English UI, content auto-translated into English" and "English UI, German content in German, other languages auto-translated into English". Also, it's impossible to express different preferences for UI and content.
The situation for any site that offers search over a dataset in multiple languages is even weirder: should Accept-Language cause search to exclude or downrank some results?
Many people (in particular people learning a language) would want to be able to control all of these aspects independently, which wouldn't be possible if everyone took heed of Accept-Language.
Fedi instances usually use only a single URL for a post. If you request application/ld+json, as ActivityPub spec mandates, you get the ActivityStreams object that corresponds to the post. If you specify nothing or request text/html, you get a web UI for that post.
The advantage here is that regardless of the client you use to view a post (an ordinary browser or an ActivityPub client) the URL the post is identified with is the same.
Of course, we could achieve a similar effect by mandating that the canonical URL return an html page that has a meta tag/header/... that points at the URL where one can retrieve the ActivityStreams object, and vice versa. I would consider that to be much more complicated though.
@Ange One thing that's not apparent from the diagram is that DICOM can store weirdly-dimensional things: it's used to store 3d images (e.g. sequences of 2d slices in MRI) and reportedly is also used for animated 2d and 3d images. I'm not sure if higher dimensionalities are possible and, if so, used in practice.
I also recall that it supported large bit depths (because e.g. that's very useful for X-ray images), but that is less surprising.
robryk
boosted
A DICOM image.
DICOM is the medical image format, very TIFF-like.
Its preamble makes many polyglots possible, including TIFF-DICOM sharing the same image data.
It doesn't tolerate appended data - but just append a "private" tag if needed.
@8petros I'm sorry for being brusque(or something similar; not sure what exactly) and thank you for directness.
That seems to talk about adding it in quantities that cause a person to ingest hundreds of mg of it per day. If you had even a 10g heatpack, you'd need to split it across ~50persondays to be in the same dose ballpark.
The reference for it being considered generally safe in the US lists sub-1% accepted levels in various foods: https://www.accessdata.fda.gov/scripts/cdrh/cfdocs/cfcfr/CFRSearch.cfm?fr=184.1193
@8petros I would appreciate if you elaborated on the cost/failure point increase (either I don't see something obvious -- it seems to me that everything should be ~comparable -- or I failed to get the difference across).
Unrelatedly, I noticed:
> The solution that remains after the reaction is completely harmless and can be reused as a food additive
or antifreeze liquid.
~1/4 of that liquid is calcium. Daily recommended intake of calcium is on the order of magnitude of 1g. Random sites on the internet claim that taking more than ~3g per day is a bad idea and NHS claims that it can cause diarrhea (https://www.nhs.uk/conditions/vitamins-and-minerals/calcium/). It might be a bad idea to drink that.
Have you tried a reverse variant of A, where the inner bag contains solid cacl2 (and, to make breaking it easier, some air or some mechanical breaking aid)? If it works it might be safer than A: you will never get solid cacl2 leaking out.
I enjoy things around information theory (and data compression), complexity theory (and cryptography), read hard scifi, currently work on weird ML (we'll see how it goes), am somewhat literal minded and have approximate knowledge of random things. I like when statements have truth values, and when things can be described simply (which is not exactly the same as shortly) and yet have interesting properties.
I live in the largest city of Switzerland (and yet have cow and sheep pastures and a swimmable lake within a few hundred meters of my place :)). I speak Polish, English, German, and can understand simple Swiss German and French.
If in doubt, please err on the side of being direct with me. I very much appreciate when people tell me that I'm being inaccurate. I think that satisfying people's curiosity is the most important thing I could be doing (and usually enjoy doing it). I am normally terse in my writing and would appreciate requests to verbosify.
I appreciate it if my grammar or style is corrected (in any of the languages I use here).
Joined Nov 2020
