Open Bar: Curbside Pickup Only

[Scribbly]

December 2019 I was hospitalised with a mystery respiratory illness which did not respond to the usual treatment. Blood oxygen was frighteningly low. I could not taste, smell or *hear* which is a symptom I haven't heard discussed much but was easily the most distressing for me other than... y'know... not being able to breathe. I spent the better part of two weeks in and out of consciousness and was pretty sure I was on my way out.

I was eventually discharged and was still very ill until January. My hearing didn't return fully until March. I suffered bouts of sudden exhaustion, fatigue and nausea until June and I still sometimes get exhausted though generally when I've exerted significantly. It has made it difficult to get back to stuff like regular exercise and I'm hoping that as I do I'll discover a lot of it is down to 2-3 years of minimal activity rather than anything permanent.

There were no tests so I'll never know for sure. But I'd been in and out of London a lot around that time and I suspect that Covid was with us during that flu season and people just didn't know what to look for.

[Cain]

It then took a month to get my energy levels back to approaching normal. I am fairly certain it was Covid, despite the test coming back negative.

There are definitely false negatives. I know several people who've recently had a horribly bad cold that totally wasn't covid because an LFT was negative.

My partner and I both got it back in March or April 2020, shortly after it officially came to the UK. There weren't any vaccines or self tests. Masks were in short supply, and it wasn't yet clear if it was a good idea to buy them or to leave them so there'd be enough left for the nurses. I got sent to a hospital and made to wait in a chalk square outside, then someone in head-to-toe PPE asked me some questions and watched my breathing, without any physical contact. It's weird to remember it being taken that seriously. They said I'd probably had it and we should both isolate for a week. I don't remember much from that week except being absolutely fucking knackered. I'm also pretty sure it made me less mentally sharp for months. No fun.

We're both up to three or four jabs now, I've lost count.

Yeah, I'm fairly convinced that I'm in the 8% (I think it was 92% accurate) that got a false answer back. I've had flu before and it didn't feel like that, with the recovery. Usually with that, once the fever breaks you start to feel much better, quite rapidly (you're still a wreck, but you're an improving wreck). And the long recovery fits Covid too well.

[Suu (parody account)]

December 2019 I was hospitalised with a mystery respiratory illness which did not respond to the usual treatment. Blood oxygen was frighteningly low. I could not taste, smell or *hear* which is a symptom I haven't heard discussed much but was easily the most distressing for me other than... y'know... not being able to breathe. I spent the better part of two weeks in and out of consciousness and was pretty sure I was on my way out.

I was eventually discharged and was still very ill until January. My hearing didn't return fully until March. I suffered bouts of sudden exhaustion, fatigue and nausea until June and I still sometimes get exhausted though generally when I've exerted significantly. It has made it difficult to get back to stuff like regular exercise and I'm hoping that as I do I'll discover a lot of it is down to 2-3 years of minimal activity rather than anything permanent.

There were no tests so I'll never know for sure. But I'd been in and out of London a lot around that time and I suspect that Covid was with us during that flu season and people just didn't know what to look for.

The first "official" Covid case in Florida was recorded on January 1st 2020, it was a child in Jacksonville, but because it was a minor, the records were not released for way too long because "There's no way this is COVID, it has to be another coronavirus!" My sister was very, very ill for Christmas that year, and then gave it to my husband. I've never seen him so sick. It didn't stick with me, I had 2 days of "meh" and then a roaring, deep hot cough for a month. We were told it was RSV because all of our flu tests popped negative. Sister worked for a Scientologist veterinarian at the time but my dad also travels everywhere, including Italy. So she's sure she got it from a Scieno that just returned from China who brought in their pet to her work, or Dad brought it home from Italy where it already was, also. They found it in waste water going back as early as November in some cities. It was spreading before they caught it.

Remember that mysterious "vaping disease" in Fall of 2019? Yeah, that.

[Cain]

A true Scientologist would simply be able to redirect matter to avoid catching Covid.

Report them to the nearest OSA official.

[Anna Mae Bollocks]

I mean, most of Tumblr sucks, absolutely. But it's still going to be an easier entry to the fediverse than Mastodon, since you know exactly what you're getting, exactly how to get it, and you aren't going to be effectively banned for breaking bizarre rules that nowhere else online has that are buried in a rules document that's functionally closer to a ToS in size and impenetrability than a list of rules.

Spoutible looks promising, at least in theory. The Bot Sentinel guy is building it.
Not getting my hopes up, everything has a way of turning to shit. And the word "safe" is kind of off-putting. But it makes a lot more sense than Mastodon.
https://spoutible.com/

[altered]

There's many promising looking possibilities, but if they can't get market share they'll be also-rans at best. Mastodon has the benefit of bright-eyed idealists and tech-fetishists that guarantee it has and will continue to have a market share, the same way IRC is still going even now that we have a thousand mildly improved options for chatting with people.

These other projects have nothing right now, and most of them will fail and probably wind up having their user data used for credential stuffing attacks.



In other news, I saw some people freaking out about AI coding. I'm a unique beast, in that I know how computers work but cannot write code due to some highly specific dyslexia -- I figured I had to give it a try. Suffice to say, the state of the art there is more along the lines of HackerTyper than anything useful. At best it will be a tedium-remover, since it can't even produce working Python code without involving debugging that my dyslexia crashes ashore on. Fucking about with wording can get you some results (I almost hacked my way to something that could produce an error further than ten lines down!), but you have to already understand computers well enough to make it irrelevant for most everyone BUT me, and it's useless to me because I can't debug it. You can't stick your manager in front of OpenAI's Codex and have him produce anything much more complicated than Hello, World. Coders aren't out of a job. The AI Art Thing this ain't.

[chaotic neutral observer]

You can't stick your manager in front of OpenAI's Codex and have him produce anything much more complicated than Hello, World. Coders aren't out of a job. The AI Art Thing this ain't.

I asked that OpenAI thingy three questions.

The first was a toy problem, but in a difficult and somewhat obscure language.  It gave me a pretty good answer; I was quite surprised.  (I wanted a gray coder written in VHDL).

The second was a highly mathematical problem in a mature and well-understood problem space, in a common language.  It thought to itself for a while, and then failed with an internal error.  (I wanted a Reed-Solomon decoder written in C).

For my third question, I asked it to provide an algorithm for a very well-known problem, which is generally thought to be unsolvable (the experts in the field believe it can't be done, but haven't yet managed to prove it).  The provided answer was utter nonsense, that would have earned a failing mark in any second-year computer science class.  (I wanted an algorithm to solve the travelling salesman problem in polynomial time).

So, no, I don't think that anyone's job is in danger quite yet.

[altered]

I've been painstakingly describing step-by-step how to parse a particular CSV, transform it into a tree, and do some extra work to make an SVG out of the tree. (This is a project I need to have someone do for me anyway, sooner rather than later. Figured it made a good starting point for testing Codex since I have a very detailed writeup.)

It can't do it.

ETA: I can get it to make surprisingly complex output that looks right when I can read it at all, sometimes. I managed it once with Python, and once (with more difficulty reading anything) with JS. But I can't make it do so consistently (more common is 80% looking good and 20% extremely wrong -- again, out of what I can read), and the output isn't functional.

[Suu (parody account)]

A true Scientologist would simply be able to redirect matter to avoid catching Covid.

Report them to the nearest OSA official.

My sister is not a Scientologist, that was clearly the problem. Regardless if they paid her double time to attend "religious acceptance training" at her job. Wherein she would bank close to $40 an hour sitting there making fun of shit over the course of 2-3 days.

[Scribbly]

I've been painstakingly describing step-by-step how to parse a particular CSV, transform it into a tree, and do some extra work to make an SVG out of the tree. (This is a project I need to have someone do for me anyway, sooner rather than later. Figured it made a good starting point for testing Codex since I have a very detailed writeup.)

It can't do it.

ETA: I can get it to make surprisingly complex output that looks right when I can read it at all, sometimes. I managed it once with Python, and once (with more difficulty reading anything) with JS. But I can't make it do so consistently (more common is 80% looking good and 20% extremely wrong -- again, out of what I can read), and the output isn't functional.

Reminds me of the person who got it to generate a paper on a relatively obscure area of study - something to do with geology if I remember right.

It looked good. Convincing, even. Had a lot of references which used the right researcher names.

Only issue was the actual papers cited didn't exist and whilst the right jargon was used in the right order the actual meaning was nonsense.

[altered]

Funny enough, I think academia is going to be the first place to get well and TRULY fucked by AI for the combination of 4 important aspects:

1: Academic papers are usually kind of samey (with good reason), which makes writing them very automatable.
2: Feeding source material and new research in and getting a paper out is something most academics who are worth a damn think they would enjoy doing, because it'd reduce the time spent on writing papers and lower the barrier to entry for getting helpers on board.
3: The bleeding edge of AI is largely coming out of academia, so will likely have a lot of access to academic source material.
4: AI models that have high contextual awareness are the norm now, which means you just have to turn the temperature down real low and have it focus on turning data into words with zero creativity.

We already have shit like computer aided chemical synthesis design and automated theorem provers out there literally churning out almost wholly computer generated papers, and have had them for years now. Academia is already primed for this sort of thing and people would hardly notice at first, even in the "hard sciences". It might even have good side effects. Then again, it'll probably concentrate institutional power even more with good politicians who can't do science worth a fuck.

[Cain]

Meanwhile I'm using the AI to shitpost with my RP group:

[altered]

A couple hours of work was put into this data visualization shit, trying to hand-build a map again. The idea was, if I just find a set of equations to plug numbers into, I can shut up and calculate everything. I knew one of the hazards of this would be a lot of blank space, but NOTHING prepared me for this bullshit.

Behold: ONE subsystem. https://cdn.discordapp.com/attachments/1002305555207041065/1063126320491528293/FmR9ZJNXwAEzNGw.png

How many do we have? Why, 7! 8 if you count the exoselves and 9 if you count the Pathways. Kill me. I crave release.

Of those very tiny circles in the group of seven, one reads "Dulug-3". Another reads "Adrusu". Can you tell which goes where? This is, admittedly, heavily zoomed out, but the zoom was needed to have any fucking hope of showing the whole thing. Why? 10000 pixels wide. It's impossible to make this usable as a reference to this fucking system.

So I'm down to trying to figure out how to make something interactive, because if it ain't interactive it ain't fucking happening. Good fucking god.

[Scribbly]

I think I've just been banned from the local 'find roleplay groups' facebook group for ... posting an ad for my game?

Currently pondering what level of revenge to take if this turns out to be correct.


ETA: It has now been confirmed.

Apparently I am not well-known enough as part of the community to be advertising my games, and the person who confirmed this got some static from them for doing so.

Excellent.

[Suu (parody account)]

Funny enough, I think academia is going to be the first place to get well and TRULY fucked by AI for the combination of 4 important aspects:

1: Academic papers are usually kind of samey (with good reason), which makes writing them very automatable.
2: Feeding source material and new research in and getting a paper out is something most academics who are worth a damn think they would enjoy doing, because it'd reduce the time spent on writing papers and lower the barrier to entry for getting helpers on board.
3: The bleeding edge of AI is largely coming out of academia, so will likely have a lot of access to academic source material.
4: AI models that have high contextual awareness are the norm now, which means you just have to turn the temperature down real low and have it focus on turning data into words with zero creativity.

We already have shit like computer aided chemical synthesis design and automated theorem provers out there literally churning out almost wholly computer generated papers, and have had them for years now. Academia is already primed for this sort of thing and people would hardly notice at first, even in the "hard sciences". It might even have good side effects. Then again, it'll probably concentrate institutional power even more with good politicians who can't do science worth a fuck.

We already have software to catch them, and honestly, some of them are very, very "off" when you read them. It's a regurgitation of facts but with no argument. It's fucking WEIRD. That, and they cite incorrect 99% of the time. I know all of this will be fixed eventually and I'm out of a job to a bunch of Zoomers who speak entire sentences with one word, but it's not making the waves people thought it was going to.

Also, I'm in the humanities, so our papers are generally more flowery and have a detectable voice. I'm sure dry STEM professors with no sense of critical thinking skills will think their students are geniuses or some shit reading this garbage.