Jonathan Lamothe reshared this.

mastodon - Link to source
👏 Poison 👏 your 👏 data ☠️
Jonathan Lamothe reshared this.

mastodon - Link to source
The media in this post is not displayed to visitors. To view it, please go to the original post.

🌌 🎶 "Time And Space", the new album from LAD was released yesterday and I am chilling out to it now.

You can get your copy of this deep space ambient masterpiece here:

laddadoane.bandcamp.com/album/…

You'll be glad you did. And yeah, there's a lot more where that came from.

#music #ambient #space

Time and Space, by Ladd Allen Doane

9 track album
Ladd Allen Doane
#space #music #ambient

reshared this

Jonathan Lamothe reshared this.

mastodon - Link to source

LLM access is relatively cheap now because the LLM vendors are discounting their price at a massive loss, subsidized by VC, in order to get you addicted and to drive as much skilled human labor as possible out of the workforce permanently.

The goal is monopolization, and if they’re successful, you’ll see monopolistic pricing in the future.

in reply to Jonathan Lamothe

mastodon - Link to source

Boyd Stephen Smith Jr.

@me Generating test data, as a complement to QuickCheck/SmallCheck generators. I think LLMs might "explore the probability space" in different ways than manually written generators. But, I haven't validated this in practice.

I've been fairly disappointed with LLMs output all the times I've tried them. Too many hallucinations around factual data. Too little... variety(?) when doing fiction. The image generators seem better than me, but I have declined to use them (much)because I assume the image generators are "stealing" from the recognition/attribution of artists that make their art publicly visible. I know the code generators "steal" copyleft code, most likely including mine.

I don't like saying LLMs capabilities are bad, because I don't use them, for ethical reasons, enough to really know what their current capabilities are.

@Jonathan Lamothe

friendica (DFRN) - Link to source

I'm getting closer and closer to passing the entrance exam for this job. I also learned a little tidbit about why they're always hiring: apparently, "AI-generated" transcripts are inadmissible in US courts.* As much as they might like to, they legally can't replace this job with AI.

Combine that with the very small overlap between people capable of passing this exam and people actually willing to jump through those hurdles, and you have glut of available work.

* At least for now. Give the techbro billionaire class time to keep eroding the US legal system, and who knows?

Jonathan Lamothe reshared this.

mastodon - Link to source

(Following thread was prompted by people pointing out that the Bluesky dev team seems heavily into vibe-coding now and originally posted on said vibe-coded Bluesky platform that is now constantly failing.)

Over the past year, every single time one of the apps or services I use suddenly became less reliable and more buggy, I never have to look far for the "Claude is amazing and now writes most of my code" post for the devs involved.

reshared this

in reply to Baldur Bjarnason

mastodon - Link to source

Elias Mårtenson

I always enjoyed Universe Today, but once you get llm psychosis, everything becomes possible.

youtube.com/watch?v=vkhZHR_hs4…

at the 47 minute mark it really goes off the rails.

SpaceX's AI Data Centres Might Actually Be A Good Idea. Here's Why

Auf YouTube findest du die angesagtesten Videos und Tracks. Außerdem kannst du eigene Inhalte hochladen und mit Freunden oder gleich der ganzen Welt teilen.
Fraser Cain (YouTube)
Jonathan Lamothe reshared this.

mastodon - Link to source

@screwlisp is having some site connectivity problems so asked me to remind everyone that we'll be on the anonradio forum at the top of the hour (a bit less than ten minutes hence) for those who like that kind of thing:

anonradio.net:8443/anonradio

He'll also be monitoring LambdaMOO at "telnet lambda.moo.mud.org 8888" for those who do that kind of thing. there are also emacs clients you should get if you're REALLY using telnet.

Topic for today, I'm told, may include the climate, the war, the oil price hikes, some rambles I've recently posted on CLIM, and the book by @cdegroot called The Genius of Lisp, which we'll also revisit again next week.

cc @ramin_hal9001

#LispyGopher #Gopher #Lisp #CommonLisp

#lisp #gopher #commonlisp #lispygopher @Ramin Honary @Cees de Groot @screwlisp
This entry was edited (2 weeks ago)

reshared this

in reply to Michael T Babcock

mastodon - Link to source

John Abbe (aka Slow)

Geeks under a certain age are impressed by the idea one was messing about in massively multiplayer worlds in the 1980s. It was early!

I ran into TinyMUD first, and via TinyMUCK their Forth-based MUD language, MUF. Something about programming in MUDs lent itself to thinking in objects though and thinking about the things I wished I could do I (later realized I) started reverse-engineering object-oriented coding.

(I'd had earlier encounters w LISP, so at some point I realized what I was doing.)

Jonathan Lamothe reshared this.

mastodon - Link to source

A follow-on to my "Nazi Sucker-punch Problem" post, to address the most common argument I get, which boils down to:

"""
Moderated registration won't stop Nazis, because they'll just pretend to be human to fool moderators, but it will stop normal people, who won't spend the effort to answer the application question or want to wait for approval.
"""

Okay, I'm going to try to use points that I hope are pretty acceptable to anyone arguing in good faith, and I'm going to expand the definition of Nazis to "attackers" and lump in bigots, trolls, scammers, spammers, etc. who use similar tactics.

Attackers: we can group attackers into two main types: dedicated and opportunistic. Dedicated attackers have a target picked and a personal motive—they hunt. Opportunistic attackers have an inclination and will attack if a target presents itself—they're scavengers. In my years of experience as an admin on multiple Fedi servers, most attackers are opportunistic.

Victims: when someone is attacked, they (and people like them) will be less likely to return to the place they were attacked.

In general: without a motive to expend more effort, humans will typically make decisions that offer the best perceived effort-to-reward ratio in the short-term (the same is true of risk-to-reward).

Why does any of this matter?

Because it all comes down to a fairly simple equation for the attackers: effort > reward. If this is true, then the opportunistic attackers will go elsewhere. If it isn't true, then their victims will go elsewhere.

How can we tip that scale out of the attackers' favor?

By making sure moderation efforts scale faster against attackers' behaviors than against normal users' behaviors.

- A normal user only has to register once, while an attacker has to re-register every time they get suspended.

- A normal user proves their normality with each action they take, while every action an attacker takes risks exposing them to moderation.

- A new user / attacker likely spends a minute or two signing up, while a moderator can review most applications in a matter of seconds. Yes, attackers can automate signups to reduce that effort (and some do, and we have tools to address some of that, but again, most attackers aren't dedicated).

- Reviewing an application is lower effort than trying to fix the damage from an attack. As someone who gets targeted regularly by attackers from open-registration servers, I'd personally rather skim and reject a page-long AI-generated application, than spend another therapy session exploring the trauma of being sent execution videos.

I believe this points to moderated registration being the lowest effort remedy for the problem of the Nazi Sucker-punch. So before we "engineer a new solution" that doesn't yet exist, we should exhaust the tools that are already available on the platform today. Yes, we could implement rate limits, or shadow bans, or trust networks, or quarantine servers, but we don't have those today, and even if we did, there's no evidence that those would be a better solution for Fedi than moderated signups.

Will it stop *all* the attackers? No. But it will stop most opportunistic attackers.

Will it deter *some* potential new users? Yes. But communities are defined by who stays, not by how many come through the door.

lgbtqia.space/@alice/115499829…

🅰🅻🅸🅲🅴 (🌈🦄) (@alice@lgbtqia.space)

Why reactive moderation isn't going to cut it, aka, "The Sucker-punch Problem". Imagine you invite your friend—let's call him Mark—to a club with you. It's open-door, which is cool, because you like when a lot of folx show up.
🅰🅻🅸🅲🅴 (🌈🦄) (LGBTQIA.Space)
Unknown parent

mastodon - Link to source

Yet another Josh

True but the fash will play along and be nicey nicey. Bouncer says yes. Wash rinse repeat, ala nazi bar.

Whereas if theyre identified as fash, their account isnt deleted.

Its just reduced engagement to other fash. Most dont catch it. They get bored cause nobody interacts, and they go away for real, without spamming account after accuont.

friendica (DFRN) - Link to source

So, there was a post on the fedi about a project Johnny Harris was working on. Some people in that thread seemed to think that he was untrustworthy, even going so far as to posit that he might be a CIA asset. I had no idea why they believed this, but it was echoed by more than one person.

I am familiar with Johnny's work. He always seems to do a good job of citing his sources (at least to my casual inspection). I asked about this distrust but received no response. Perhaps they thought I was sealioning?

So, I'm asking here: Is there an actual valid reason to distrust him that I'm simply not aware of, or is just stemming from the fact that he likes to shine light on things that some would rather not have light shined on?

Jonathan Lamothe reshared this.

mastodon - Link to source
The media in this post is not displayed to visitors. To view it, please go to the original post.

If a Klein bottle could wear pants, would it be like this or like this?

#mathstodon #math #maths #shitpost

#math #maths #shitpost #mathstodon
Jonathan Lamothe reshared this.

mastodon - Link to source

I don't think programmers and sysadmins get how much there is to learn and how intimidating it is for normal people to host their own software.

For one most of us don't have a computer that is running 24/7, which means we need to rent a server which we have no idea how to go about doing.

And then there's an entire arcane art to running software that can speak to the internet without your server being taken over and used to send spam to half the planet

This entry was edited (1 month ago)

reshared this

Jonathan Lamothe reshared this.

friendica (DFRN) - Link to source

Hey Fedi,

For those who don't know, my mother had a major #stroke a little over a month ago. We're very fortunate to live in a country (Canada) where we have free #healthcare, but as her discharge from the #hospital looms closer, we're having to raise funds to make #accessibility modifications to my parents' home so that she can return. Boosts are welcome (and appreciated).

gofund.me/a69e0cdc4

#a11y #MutualAid

a69e0cdc4

gofund.me
#Hospital #healthcare #accessibility #a11y #stroke #mutualaid
Jonathan Lamothe reshared this.

glitchsoc - Link to source

Him: well what's your big brain idea to eliminate food assistance fraud?

Me: universal food assistance

Him:

Me: because there wouldn't be a system to cheat, everyone would just get a check

Him: and who does that benefit?

Me: family farmers and human beings who rely on food for nutrition

Him: what about rich people?

Me: what about them?

Him: you would give them food assistance too?

Me: that's what universal means

Him: you can't do that

Me: yesterday you said I couldn't tax them and now you won't let me feed them either?

Him: they can afford food

Me: then it should be fine to tax them

Him: but if you tax them they won't have as much money

Me: I'm willing to offer universal food assistance

friendica (DFRN) - Link to source

Are there any #Lisp programmers out there who use a #ScreenReader? Given how messy Lisp can be to read without proper indentation (which I imagne wouldn't translate well on a screen reader) I can't see it is being an easy language to work in without being able to see it.

I've been thinking about a way to make an editor that lets you explore a Lisp program by walking through the forms in the program in a manner similar to the way one might navigate in a MUD. Is this a crazy idea, or one with some merit?

#lisp #screenreader
in reply to Jonathan Lamothe

mastodon - Link to source

Svante

Seemingly plain (lambda () ...) is a macro that expands to (function (lambda () ...)). #'(lambda () ...) uses a reader macro to expand to the same (function (lambda () ...)).

clhs.lisp.se/Body/m_lambda.htm

Function is a special operator that returns a function. It takes either a function name or a lambda expression. The second case is what is happening here.

clhs.lisp.se/Body/s_fn.htm#fun…

A lambda expression is a list of the symbol lambda, a lambda list, and a body.

clhs.lisp.se/Body/26_glo_l.htm…

CLHS: Macro LAMBDA

clhs.lisp.se

friendica (DFRN) - Link to source

I've been taking a bunch of tests to qualify for a transcription job. They're not easy and I need a perfect score to pass. I finally failed one of the tests but managed to pass it on the retry.

They're really picky about their style guide. Fortunately, it basically amounts to syntax rules and I've been dealing with compilers that are equally picky about syntax for decades.

It also helps that all throughout my schooling my mother worked at the local university proofreading research scientists' papers and she insisted on proofreafing all my essays too.

I never thought I'd end up being happy about that.

in reply to Jonathan Lamothe

friendica (DFRN) - Link to source

Jonathan Lamothe

So the cooldown period for me to re-take the exam is over. They want me to redo The. Whole. Damn. Thing.

Fine, whatever. It'll take me a few days to get through it all, but I don't expect to fail again.

Given how their style guide is now permanently burned into my brain, there's a tiny vindictive part of me that toyed with the idea of spinning up a competing service. That's probably more throuble than it's worth, though.

I see now why they're always hiring. I wonder how many people get to the prequalification exam and then just decide it's not worth it.

in reply to Jonathan Lamothe

friendica (DFRN) - Link to source

Jonathan Lamothe

After a closer review of the style guide, I have come to realize that the question I was repeatedly getting wrong was one that wasn't even on my radar as a candidate. I misremembered a stupid trivial detail from the guide.* I was rabbit holing on the wrong two questions, one of which I still maintain can be argued to have two correct answers, but I now know which one they're looking for.

Not looking forward to sinking another ~5 hours on this thing.

* That would have had virtually zero impact on my ability to successfully do this job.

Unknown parent

friendica (DFRN) - Link to source

Jonathan Lamothe

@🇨🇦 CleoQc 🍁🦜🧶🚐🌈 It's a legal transcription job. They do regular transcription too, but they have AI doing much of it, so they're not looking for new people there.

They're smart enough to realize that AI isn't currently sophisticated enough to properly follow the various style guides required by their legal clients. I guess they realize that if the quality of the work drops, they'll lose the contracts.

That said, I'm sure all the work I do is going to be used to try to train an AI to replace me, but that's probably true of any job at this point.

@🇨🇦 CleoQc 🍁🦜🧶🚐🌈
Jonathan Lamothe reshared this.

mastodon - Link to source
Don't do AI. Look for people selling electronics they don't need anymore and use them to make the machine. Make a god out of discarded parts, and don't let it die. Use it for whatever the hell you feel like. Give it the ability to sense its reality and start programming it not to speak, or interact with anyone, but to show people echoes of the past.

reshared this

This website uses cookies. If you continue browsing this website, you agree to the usage of cookies.

⇧