ManMachine
@max@manmachine.me
autism tests be like "do you take things too literally" and then ask 50 of the worst worded questions ever known to man
Sometimes I pair-program, or more usually pair-debug, as a teaching technique: solve a problem while screen sharing, narrating my thoughts and reasoning out loud to show the other person how to track down whatever kind of problem it is.
But it has to be on a reasonably simple problem, because narrating my thoughts coherently _also_ requires thought and effort, and that reduces the amount of brain I have available for the actual problem. For the hardest problems I'm capable of solving at all, the solving and the explaining don't both fit in my head at once, and I have no choice but to say "sorry, I need to concentrate completely on this, and once I've found the answer, _then_ I'll tell you how I did it."
This week, someone pointed out the contrast between this and the 'rubber duck' / 'cardboard programmer' effect, in which explaining out loud what's confusing you can _help_ you to figure it out yourself. I'd never noticed that contrast before.
So, some problems become easier to solve if you talk through your thoughts out loud, and others become harder. That's annoying! Now you also need the proverbial "wisdom to know the difference".
@simontatham I find that in general, I can solve things a lot better when not also having to communicate. But once I get *stuck*, explaining it to someone - which forces me to put it into coherent sentences that someone else will understand - helps a lot.
This is *not* the same as solving a problem together, though, which can also be useful. The 'rubber duck' approach isn't so much about communication (with a lot of brain cycles taken up by listening to the other person's input and processing it), but more about ordering your thoughts in such a way that someone, both your listener and yourself, can more easily make sense of it. A good human 'rubber duck' does not actually offer input (other than 'hey, I don't get that bit, can you go over it again') until you're done explaining, when 1. The 'rubber duck' stage is over and you possibly move on to pair solving, and 2. You more often than not have already found the answer.
@simontatham On a tangent, I also find that I do the 'solve together' better in written form. Probably because written text is less disruptive to my chain of thought than spoken, and because task switching between two pieces of writing or writing and thought is a less expensive operation for me than task switching between written and spoken information.
@wynke huh, that's interesting – I think I have slightly the opposite preference, because I can speak out loud without having to faff about with the computer's user interface! Changing keyboard focus into the Slack window (or similar) is more annoying to me than changing cognitive focus from keyboard to voice.
@simontatham I think you first need to have come to a point where you basically have all the pieces in your mind already to some degree before the rubber duck effect works.
It seems to work best for cases where you're stuck on something and you're going in circles, at that point if you need to write down or talk out loud about the problem to make sure you're not missing some critical step or piece of the puzzle.
In a sense it's like writing out the formula for a complex equation to find X
@simontatham have you ever done pair-programming with someone on a similar level as you, where you don't need words, they can see your thoughts just by following your cursor on the screen?
@wolf480pl no, and I'd be surprised if it worked that way, because my usual experience is that people have wildly varying styles for how they use text editors and command lines. To be able to infer someone's thoughts just from how they were moving the cursor, I'd expect they'd have to be not just at about my skill level, but more like an actual clone of me in particular!
@simontatham I guess I may've exaggerated a bit.
IIRC it's more like, you have a few ideas on what the next edit could be, you see your friend move the cursor in the direction of one of them, so you know what they're about to do, so you can provide feedback (usually silence or "yup" if they hesitate, sometimes a caveat or counterargument) even before they do it.
I've experienced that a couple times and it was a long time ago, so I might not be remembering all the details correctly.
I dunno that 'the problem becomes harder', more like with limited capacity for cognitive load, some problems will overload the solver/explainer.
This is a really important distinction: I don't know that the effect you have described makes the problem subjectively (ever-)harder to understand, in a mirror way that rubber duck debugging makes it ever-easier, and your wording seems to posit this symmetry.
@simontatham The advantage of the rubber duck is that if you explain vaguely or garbledly, it doesn't object, so you can explain clearly if it helps and vaguely if that helps :)
I think there's also the level where if the problem is *too* big to fit in my mind at once, I need to track "where I've got, which part I'm trying to understand now" in writing.
@simontatham in the first case I would describe the extra room being taken up by the needs of the other person rather than by explaining which would suggest the advantage of rubber ducking is that you don't worry about the duck's needs.
@simontatham
I just went through an interesting debugging process. Towards the end my thoughts were "why is it... Hang on... What if I..." There wasn't a sharp line where I suddenly knew, what was wrong. I guess it's in the "that's odd" family of phenomena. And demanded 100% of my brain for debugging, so that even my thoughts were incoherent.
@LiseAndreasen yes, that "answer bubbles up from your subconscious" phenomenon is very familiar, and indeed hard to narrate or even explain.
Worse still, the thing you _most_ want to explain to a student is how your subconscious solved the problem in the first place, and even you may not know that, because subconsciouses are amazingly bad at showing their working!
@simontatham With the Rubber Duck approach you can switch to inside problem-solving (and shut up) any time you need to because the Duck doesn't actually give a quack - so it can be a good technique to get you a little further down the thinking road
With a student you need to stay in communication mode which doesn't pair greatly with inside problem-solving so this will not work optimally for new problems
When this happens and you need to do inside problem-solving but you are with a student you
@simontatham If I'm reading that right, it seems like a weird way to operate..
Rubber ducking is not a teaching exercise, it's something you do when you've tried and failed to solve a problem, and you need to get help.
If it's you teaching them, it would probably better to get them to explain a problem of theirs to you..
Teaching with a live problem that you don't know the answer to is a decent recipe for failing to teach anything beyond managing frustration most of the time, I suspect.
Friends: "hey i vibecoded a thing ha ha"
Me: "that's kind of sad"
Friends: " >:-( why aren't you praising us"
I'm apparently breaking a friendship contract by not praising friends for doing stuff I find unethical 🤔 but friends don't let friends vibe code
@nina_kali_nina
They must already be used to the fawning praises of the bot about everything they do ?
Must be sad for you, because you wish to be true with your friends ?
@nina_kali_nina
My social world definitely got smaller since the proliferation of LLMs. And then in a way it got bigger again with the wonderful antislop circles on fedi. But yeah, on individual cases, seeing people you respected fall into the llm traps, it's heartbreaking.
@ozzelot @nina_kali_nina Not a day goes by that I don't think about H. G. Wells's 'Time Machine"
@thankfulmachine @ozzelot I've read it so long time ago I can't remember anything. How come the current situation reminds you of it?
@nina_kali_nina Fortunately most of my friends and acquaintances are anti-AI, but the few that aren't, whenever it comes up it gets incredibly awkward.
It never happened with any really close friend (yet), thankfully.
@nina_kali_nina The especially galling one for me is something that would, in a normal world, be done purely for fun or technical achievement. Like a tool or a demo for a dead computer system. You’ve taken a joyful activity and outsourced it to a machine. I am supposed to be impressed by this?
the adoption of generative AI within the demoscene community is something I never would have anticipated. I always viewed the demoscene as a celebration of human ingenuity.
saying this as I think if any group of people could make something actually worthwhile with AI assistance it could be demosceners, but the idea of throwing up some generative AI art in the middle of a demo and pausing on it like it has any artistic resemblance to something hand-pixelled in Deluxe Paint on an Amiga is just entirely baffling to me.
The 'AI-generated' category at Assembly Summer was also a bit of a slop show. psenough as a pretty brutal and entertaining takedown of the whole mess
@gloriouscow @nina_kali_nina Even last year I feel like it was widely frowned upon but I’m not too connected to the wider ‘scene.
I have also heard that it’s basically destroyed all the Defcon puzzles this year. As in, pumping puzzles through Claude is such a competitive advantage (even on puzzles thought to be AI-resistant), you can’t win by sincerely engaging with your brain. So there’s no enjoyment to be had there anymore.
well, it's stil divisive and as a group they aren't a monolith but there are plenty of groups and some very big names in the scene embracing it
also he mentions my emoji code stuff which is fun
speaking of ruining puzzles, someone reached out to me wanting to use my emoji code as part of an ARG they are running - they had a COM file that would print the next clue, and so the task would be to figure out what the emoji was.
but if you paste the emoji in to ChatGPT it correctly guesses it's an executable, and it even links to my blog about it.
so what would have been a fiendish puzzle is now just like copy and paste, get answer
@gloriouscow @bytex64 wow, this is kind of incredible but also kind of scary
Their capabilities have been increasing incredibly fast. If you're fiercely anti-AI and the last time you tried a SOTA model was even a year ago, you do not have have an accurate conception of what they can do.
I used to test them with various questions about retrocomputing because they were always hilariously wrong about things. they aren't anymore. they used to be abysmal at assembly language, now they can generate entire operating systems for 16-bit computers in assembly language.
thats why demosceners are starting to use them. they can be used to code-golf bytes, improve compression algorithms, including some very heavy math. they can autonomously operate emulators, step through code in debuggers, and profile performance improvements without human interaction.
i saw an example of someone asking 5.6 Sol whether a certain piece of code was faster than another and Sol wrote an entire emulator for the CPU in the middle of the chat session to evaluate that. just as a one-off.
the landscape is changing faster than society can ever hope to keep up with it.
this is not a defense of AI, in a sense it makes the problem worse.
if AI only ever produced slop, then it would inevitably be seen for what it was. The real problem is what happens when AI provides an unmistakable productivity gain for a competent operator.
when given the choice between taking an ethical stand vs something that provides a convenience , your average joe will pick convenience every time, especially when the ethical consequences are nebulous and far removed.
"the data center is not in my back yard."
"the water gets recycled."
"the stolen training material is already in the weights and complaining about it now doesn't help anyone"
this is how the whole thing becomes some sort of a footnote in history books :
in the early 21st century, training methods employed for early AIs were occasionally controverisal.
That's what your grandchildren will read.
@gloriouscow @bytex64 it's a race to the bottom, and it's always has been: (
@gloriouscow @bytex64 this is kind of interesting because at the same time the code quality it generates is pretty abysmal. Even with all the seas boiled. It's simultaneously incredible how complex the generated code could be, and how wrong or unnecessarily complex it could get at the same time.
well, the code quality will only improve. as the saying goes, this is the worst it will ever be
there is a fundamental issue with the technology in terms of context windows, which is how you get things like claude writing the same function four or five times.
if no one is reviewing 20,000 line PRs you will end up with an unmaintainable codebase. if you are asking AI to do targeted implementations of functions and providing the architecture and APIs you will probably be fine, forever.
@gloriouscow @bytex64 I am sorry but I don't buy it. I've seen the targeted changes that were output of Fable by engineers who used to be talented and it's slop, and I've seen engineerings losing a grip on their skills, many times, so they stop being able to discern what is a good output. Perhaps, many just don't care.
I also don't buy the argument that it'll only ever go up from there. If markets get disillusioned, pumping up money into the industry will stop, and no one will care to train ridiculously large models if it's not going to be feasible. On top of that, there's risks of model collapse and the lack of new training data.
I want to believe
what i'm really curious about is where the the ceiling of the LLM transformer-based architecture actually is. very smart people like Yann LeCun seem to believe LLMs will not get us to AGI and there are good reasons why, but so far just "moar parameters" seems to be quite effective with no signs of asymptotically approaching some limit
@gloriouscow @bytex64 am I allowed to reject steak even if it smells good and causes my stomach rumble? Because I know I'll hate myself for directly causing great harms that were completely unnecessary
if you found meat revolting then avoiding it wouldn't be an ethical stance, it would just be your preference. it's precisely finding it delicious and still abstaining where you demonstrate ethical agency
to wheel this back around
i see a lot of anti-AI arguments trying to convince people that AI is a dumb useless hallucination machine
and there is truth to that. but when AI has debugged that memory leak you spent the last six weeks on in and it took fifteen minutes and $3.50 in tokens to do so then you will be less receptive to future arguments about how AI is utterly useless because it now conflicts with your direct experience.
if you are trying to prevent anyone from trying AI in the first place, telling them that AI is stupid and useless may be effective. If you're trying to convince someone using AI today, you are going to sound out of touch
@gloriouscow @nina_kali_nina @bytex64
I've been working on a niche PC emulator for the past 10 years, just for the fun of it. I'm sure a machine could now create it in a couple of minutes.
This ai garbage sucked all the joy out of this. I lost communities, friends, and interest in what was a passion for me.
I just stopped programming as a hobby, it all feels so pointless now.
If people want an emulator, they can just ask claude or whatever is the chatbot du jour.
Oh well, time to find another, maybe healthier passion I guess.
@dukeboitans @nina_kali_nina @bytex64
Someone on a forum posted a vibe-coded emulator that used an SVG i had made of the IBM 5150 motherboard and built a netlist out of it and lit up all the traces, it was simulating all the digital logic on the motherboard, and somehow managed to do it in realtime
I had a moment where i thought "this is it, i have been replaced" but then i started poking around in it and there was a of suspicious shit. lines just blinking on and off with nothing tied to them, the cpu was not cycle-accurate at all despite the readme suggestion it was
Someone even forked MartyPC and fed it to Fable and asked it to add the 286 and it launched on some bone-headed plan to shove 286 support into the NEC V20 CPU core.
we are not replaceable. yet.
that guy's fancy netlist emulator? I'm sure he's already moved on to the next project
@dukeboitans @nina_kali_nina @bytex64
at times I feel married to MartyPC. My emulator has been like the primary thing I think about day to day for the past four years. I can close my eyes and mentally visualize its structure.
even if you can vibe code something, what would keep you working on it for four years after? there's no investment of time and energy.
although this might be sunk cost fallacy talking, I have worked too damn hard on it to stop now. My advantage over vibe coders is that they will never be as tenacious or invested in the results as I will be.
@gloriouscow @dukeboitans @bytex64 I think there's great value in trying to understand something in depth. The process of understanding leads to truly novel things.
@nina_kali_nina Sadness.
In the words of late Jim Morrison: "I need a brand new friend who doesn't bother me.".
It makes me sad when I can't convince friends or acquaintances to stop using the slop machines as well. It's a very isolating feeling.
It's hard to maintain friendships like that, when you look at all the harm LLMs are unleashing upon the world, and the plethora of ethical problems surrounding them, and people insist on using them anyway.
Not like they are unaware, because we told them about all the issues!
In the end I can't help but conclude they put their own self-interest ahead of the greater good. For varying reasons and to varying degrees.
For instance, some have to use it at work, and they feel like they can't quit because there's no other jobs, and they might have dependents to take care of.
But some use it in their private time as well, and in those instances, the excuse is usually because they don't have the time or skill to do the thing they were looking to do, and if you dig deeper into that, it boils down to the fact they don't value or respect the thing they are looking to accomplish.
As such, most people who use it for code, don't understand the value of writing good, maintainable code.
Even then, i find it a hard pill to swallow given that you've literally just explained to them all the harms caused by these things, and often they look at you like you're the crazy one. Sigh.
Sometimes this feels like dealing with MAGA family members all over again. Between social media, fascist propaganda, and now ai techbro lies, there's been too much divisive brainwashing going on lately, making it very hard indeed to make and keep meaningful connections with people. In the end, it's the billionaires that win out, because there's no way we're ever going to unite for a common cause if we can't stand one another. :( But on the flip side, how the fuck am I supposed to sit at the table with fascist enablers and people who'd rather see the world burn so they can ask a chatbot dumb meme questions.
I often feel hopeless and sad, and scared for what is to come.
When the world's gone collectively mad, and if sanity is defined by that which is determined by society as normal, only the insane and fools remain as beacons of reason.
Oh crap pardon the wall of text. :s
@nina_kali_nina It especially annoyed me when The Register made an article about how there were so many talks at a vintage computer event about using LLMs to vibe code and reverse engineer vintage technology
I'm glad I wasn't there as I would just up and leave, the purpose of digital archaeology is to learn and understand, so the last thing I wanna sit through is an hour about how someone didn't learn or understand anything and just issued an LLM some commands
@nina_kali_nina
mmh, I kinda feel you; I received a "Happy Birthday" pic made with Gemini and I wondered if I should tell them to please not.
Long distance
@heyheymomo me and my homies in the late 90s when we had a LAN Party and discovered this thing called "Chatrooms"
@heyheymomo
There is an event coming up in Toronto, a queer in-person MUD get together. Like, a bunch of LGBT+ people in a big room, playing a MUD together.
@heyheymomo
I sent this to my friend who moved away, we're 13 hours apart now so my midnight is her afternoon.
it's a ... string... representing an ... integer... of an enum?
ok. sure. right.
New (old) toy: 5th gen black iPod video loaded it up with 788 tracks to go alongside my two nanos. It's modded to 256GB
#iPod #iPodVideo #iPod5thGen #Apple #VintageApple #RetroTech
me when i dont have to gaf abt whatever mastodon is doing because i refuse to run ruby on rails on my servers
@fiore i made the blunder of joining fedi without purely selfhosting
RE: https://mastodon.mit.edu/@amsomniac/117113957359596534
An off switch for "AI" is not enough. *An off switch for "AI" is never enough.* The "AI" has to be actually physically removed from the binary. Any software project who believes "Artificial Intelligence" is a real thing does not have the institutional competence to implement an off switch
dungeons in You Have Died should be
| designed by hand, same every time: | 20 |
| fully procedurally generated: | 17 |
| rooms designed by hand, but arranged differently: | 80 |
| other / results: | 5 |
Closed
@eniko I am okay with any, as long as its not made by genAi because I am not going to play your game then.
@eniko it's a good question hiding plenty, in my opinion.
If level are simply space filling corridor, this force the player to farm exp (or flee a lot).
Lot of skippable area can let the player be underleveled if they choose to or farm too much and be overleveled (unless mitigated like kess exp or less encounter).
if they need to explore and find the next stair, if they don't farm they'll have random exp. This can be mitigated by numerous floor but 🤷♀️
@eniko I'm struggling to answer because each has its benefits. If it's all handcrafted, there can be meaningfully rewarding secrets. If it's all proc gen, it makes for the most replayability since you never know what to expect. If it's hand crafted sections stitched together randomly, you get a jack of all trades that can cover both styles, but tends to lack enough for either style to be fully satisfied. Think Spelunky is the only one that had made room + random uniquely satisfying.
@eniko I'm a big fan of physical spaces that can be learned & remembered, but the content within can be surprising / changed / procgen.
obvs, do whatever you feel is best, but I'm a big fan of composable hierarchies. Handmade rooms + hand designed layers of valid combinations w/ some randomness. Think like a layered wave function collapse. You get something that feels very handcrafted but provides a great deal of variety.
random poll, do you talk to yourself out loud? like when you are alone?
boost to distract me!
| yes: | 2465 |
| no: | 564 |
Closed
do you remember when all of the techbros in the hacker news set were weirdly obsessed with bland meal replacement shakes being more efficient and the future of nutrition and they all kept shouting about how eating food with flavor and texture is irrational
and then it turned out most people like eating normal meals, and there’s well-known health consequences for all-liquid diets, and the computer science weirdo running the thing was really fucking bad at food safety, and the scientifically proven complete nutrition had gigantic gaps in the nutrition it provided, and like meal replacement shakes did in fact already exist and you can just buy them
anyway it’s good the tech industry learned from that and didn’t try to push flavorless slop that’s unfit for its advertised purpose onto a public that doesn’t want flavorless slop
I cannot express how big of a failure it is to push a full liquid diet and not know that it fucks up your mouth
you can and really should ask a doctor and then listen to the doctor
but without the bizarre claims and the puffery and the marketing, the shake is actually a really boring sub-standard product for its category
they couldn’t make it taste any good so they did a fucking Kellogg’s and made it bland as shit and decided that was good
and then they did the other half of the Kellogg’s and compromised on the supposed point of their product because it wasn’t very successful
someone decided to post on this thread and do the literal thing where they made a bunch of bizarre, incorrect assumptions about how a liquid-only diet can’t possibly be bad for the human jaw aka one of the things this thread is about
anyway speaking of tech people making assumptions about medicine, here’s Rob Rhinehart the creator of Soylent killing his gut fauna so he wouldn’t poop for a day: https://archive.ph/n03UC
> Flush toilets use enormous quantities of water so I needed a way to make it unnecessary.
> Feces are almost entirely deceased gut bacteria and water. I massacred my gut bacteria the day before by consuming a DIY Soylent version with no fiber and taking 500mg of Rifaximin, an antibiotic with poor bioavailability, meaning it stays in your gut and kills bacteria. Soylent’s microbiome consultant advised that this is a terrible idea so I do not recommend it. However, it worked. Throughout the challenge I did not defecate.
The root of the performance issues of NetBSD is uncovered: crappy BIOS limits the CPU frequency to 1400 MHz. Fresh Linux with intel_pstate can bypass it and get the frequency up to 4200 MHz, but Linux using legacy ACPI and NetBSD 11 are both capped at 1400 MHz. The internets mention that old BIOS versions used to allow higher frequencies by default, but the last version that could do that is below the version I can downgrade to. I guess reading acpi driver code it is... When there's energy
@nina_kali_nina
That is to say, how awful. Why do manufacturers do things like that?
@rodneylives apparently this update also locks the CPU frequency to 1400 MHz if the BIOS thinks the battery health is in "fair" (not even "bad") state. So I think they do it for the profit reasons, mostly
@nina_kali_nina @rodneylives Ouchy.
I'm currently running Linux on an old Thinkpad that doesn't do batteries any more (until I manage to get it fixed) and it took months to figure out how to stop the CPU being throttled using https://man7.org/linux/man-pages/man4/msr.4.html
Maybe there's a NetBSD equivalent?
@conniptions @rodneylives I tried the MSR, unfortunately the controller rewrites it immediately. It's really cursed.
@nina_kali_nina @rodneylives Double and triple ouch :(
@conniptions @rodneylives yeah :( it seems MSR used to work in an old BIOS, but not anymore. I think it's really "upwards and onwards" here
We're planning on doing a live editing session with the team. We'd love for you to join us, and chat. Do you have a preference on tech/structure/format?
| Private Zoom meeting: | 3 |
| YouTube Livestream: | 49 |
Do Not Talk To Me About "AI"
I had some cards printed up.
These are unlikely to change any minds, but they will defintely shorten some incredibly tedious conversations at the bar.
Here's a PDF if you'd like some of your own.
https://jwz.org/b/yk-o
@jwz "My Do Not Talk To Me About AI card has people asking me a lot of questions already answered by my card"
@jwz yes but local LLMs on your own computer is not as bad. the models were already trained so the worst possible crime you commit is adding a +1 on that hugging face download counter.
@curltoes "Fuck you" he explained.
@jwz i don't understand what your problem is with it i wasnt hostile towards you and you are just being mean for no reason
@curltoes 1) social media, and yes, mastodon as well, has 2 sides, ONLY 2 sides, and no middle ground - a "lets take a reasonable stance here" would not get boosted by the hype and engagement machine 2) you're engaging with someone wearing a MAGA hat, this one just says NO AI, so you gotta manage your expectations here..
@jwz Probably should specify wtf you mean by ai.
@TimWardCam @twigpi "And also fuck you", he elaborated.
@jwz I've run an offline llm on my computer. I've programmed response logic for enemies in games and played chess and other games aganst a computer. I've used text and image recognition and benefitted from an image cropping tool guessing where I wanted to crop to.
@twigpi Can you point to the place in this post that made you think that I would be even remotely interested in this anecdote?
*plonk*
@jwz can I get this on a tshirt (to go with my "please stop putting AI in everything, everybody hates it nobody asked for it" tshirt)
@jwz is it so tedious? Everyone hates it, yes, but humans are herd animals, we need common things to talk about. Weather, ai, russians...
@jwz
I don't get it. Is this like a big chunk of your identity or something?
@jwz I love it. Now I need to translate it to Spanish
@CorioPsicologia
Hey, I had the same idea! Please share if you do.
@jwz
@CorioPsicologia
A Spanish version is now available here:
https://blog.informaticalibremx.net/files/aicards_es.pdf
@jwz
@SeiriosB @CorioPsicologia Thank you for your service!
Though... I ran this past a native Spanish speaker and we question your translations of "cringe" and "motherfuckers".
@jwz
Yeah, other translations are possible for sure, but I think the overall message gets through.
@CorioPsicologia
@jwz @SeiriosB @CorioPsicologia i would be very interested in the details of this discussion! Always good to be up to date on foreign swearwords
@jwz There are an absolutely incredible number of blockables in the replies to this post, amazing really given the point about consent.
#ALT4you A table with red colored stickers with white colored text : Do not talk to me about "AI"
The backside of the stickers show : There is no ethical use of AI and The machine is digusting and must be destroyed.
@jwz *chef's kiss*
Need to cook up some or my own that say things along the line of "if I ever had respect for you, that has past", or "you having goven me carte blance to rant about AI at you for 30 minutes straight", or "you sound gullible, and if you don't think AI is a bubble, I have some tulip bulb futures to srll you"
@jwz It's like demonizing MySQL app, maybe it runs at some datacenter, and it can do video editing too. It's actually not much useful application if you notice it just takes wiki pages and
google links at face value.
@jwz Ai's are going to become the biggest waste ever. China will be making a box versions of them all within 20 years, when particle memory is achieved.
@jwz The replies to posts stating that AI stans are tedious motherfuckers invariably demonstrate that AI stans are, indeed, tedious motherfuckers.
@jwz May I suggest an alt ? Adding it to the image would make it accessible to everyone.
Two stacks of red cards with white writing.
On the front : Do not talk to me about "AI"
In a smaller font : you make me long for the sweet embrace of death.On the back :
By using AI, you have given a pass to :
- The staggering climate and ecological impact
- The outright attack on labor by capital
- The systemic disregard of consent
- The basis in wholesale theft and plagiarism
- The assault on the very concept of objective truth
- The enrichment of the most cringe billionaire fascist motherfuckers the world has ever seen
There is no ethical use of AI
If you ignore all of that, you are trying to negotiate what level of harm to others you find acceptable for your personal, selfish benefit. You advance the agenda of these hollow-eyed, joyless money-ghouls who resent art, creativity, and life itself.
The machine is disgusting and must be destroyed.
Edit as this prick blocked me because this snowflake was offended I tried to help : this was the work of my brain and fingers : no AI nor OCR were used to write this text.
@orange_lux You may not. *plonk*
@jwz you do realize the alt text they suggested was written by them, not an LLM? This is some weird knee-jerk aggressiveness to someone who took the time to try to make your post understandable to more people.
@kincat If you don't like how I write my posts, the block button is right there. I did not ask for, nor do I want, "help" from the Mastodon HOA Scolds telling me I'm Doing It Wrong.
@jwz "mastodon HOA", the expression used around here to insult the people who have done antiracist and adjacent work on this network like creating the fediblock tag?
Either you don't know the history here, and I invite you to read up on it, or you do and I wish you the day you deserve.
@jwz@mastodon.social @orange_lux
I feel so bad for boosting and even thanking you for your post. If I had noticed it was missing alt text, I wouldn’t have boosted it in the first place.
And now that I know you’re the kind of asshole who punches down on disabled people, I feel even worse, and the least I can do is block you.
@jwz @orange_lux the disabled and those who are a slow internet speed say thank you for the efforts you’re making to improve accessibility (no)
@jwz Your anti-alt-text post seems to be primarily about how it's redundant with previews. But this one doesn't have a preview, at least on my client. Am I missing something?
@jwz I liked your initial post but that’s not a good look, accessibility is lovely, this response is not
@unluckyhaircut If you don't like how I format my posts, the block button, as a I repeatedly advise, is RIGHT THERE. Here, let me help you.
@jwz @orange_lux i kinda always had my doubts on you given your tone in various posts but yeah checks out, you're just a bona fide asshole.
(and i hope as many people pile onto you as possible given how you're responding to this.)
As with any post that escapes containment, the race is now on to see who can be the most completely insufferable:
• Butthurt Slopfondlers
• Mastodon HOA Alt Text Scolds https://jwz.org/b/ykho
@jwz Your picture of some cards seems to have caused a lot of people to do the thing that was explicitly forbidden by the cards
@jwz
»meat is murder« is closer to saying to »don't be a nazi« as most farmed pigs are sent to gaschambers that cannot be compared to not using alt text.
https://watchdominion.org
@jwz Oh no, this is absolutely terrible. I can’t imagine what you are going through!
Not only do people dare to kindly offer you alt text that they took the time to write, but they do not stop there: they even dare to kindly ask you to spend ten seconds of your precious time to edit your message and copy-paste it, so that everyone could benefit from it!
People are really incredibly rude these days. You were right to spend way more time writing this long blog post to explain your fight against having to make small efforts for accessibility. 😐️
@jwz Man it costs absolutely 0 bucks 0 cents to be nice but you instead decided to spend all this energy into being a smug ableist moron. Speaking on the behalf of people using a screen reader is taking the cake.
Yeah I know you'll block me, don't care, you'll have to do the thoughtwork, my instance already muted mastodon.social specifically because of people like you.
@jwz somehow these cards remind me of a guy I met on a shared trip around the corner from the DNA, around 2015ish. He was a developer who'd been forced to attend some tedious "networking" thing and had entertained himself by pitching an app that would allow people to rent extra space in their refrigerators - "The Uber for your fridge!" and then timed how long it would take for his interlocutors to figure out he was bullshitting them. Naturally, some never did.
Tonight at the bar: "How many of those cards do you have on you right now and can I have them all?"
@jwz Do you think if one were to wire those up into a full suit of plate armor, one would finally find peace?
Crazy how most of the time I'll post about, like, the ICE torture taser gloves, and people will comment, sometimes their jokes won't land but clearly we're on the same horrified page. Then something like my "AI" post escapes confinement and these absolute lunatics show up saying taser torture gloves are good actually if you use them locally, I need to seek a moderate middle ground with the taser torture gloves, taser torture or get left behind, I don't know who the Luddites were. Wild.
@jwz Speaking of the ICE gloves… you know there’s that thing that sometimes happens at conferences where a speaker forgets they’re wearing a mic and goes to the bathroom with the mic on?
I’m wondering if we’ll hear about it the first time an ICE chud goes to take a piss with his electroshock gloves on.
@jwz we're just saying you're an asshole dude. if you can't take it that's not our problem.
@jwz extreme positions go viral in #socialmedia and #Mastodon is no exception. I am more for shades of grey since this fits my experiences better. For me a #bot creates #Alt text. My server (NRW.social) provides translations via #Deepl. Ai also helps me to use #linux and the command line. I use #birdnet and @FloraIncognita_DE to identify species for #natureconservation. Everything on the card considering bad impacts is true at the same time.
@jwz
With "AI" you mean the generative bullshit machine?
As machine learning has been used since dozens of years, and can actually do useful stuff (and not kill the planet by doing so).
@jwz I'd love the follow up from this experiment, if you care to share later. I'm intensely curious whether this is news to the average Joe, or whether they know it and choose to ignore it.
@jwz Recreated and sent to the printers thanks (can ping but can't seem to reach your website from the UK right now)
@jwz Today on "something that I occasionally deeply want to post on LinkedIn, even though I probably shouldn't."
Alt text
Two red cards, about the size of a business card, have white text.
One side of the card had very large text:
DO NOT TALK TO ME ABOUT "AI"
In small text below that:
You make me long for the sweet embrace of death
The other side of the card has smaller text:
BY USING AI, YOU HAVE GIVEN A PASS TO
• The staggering climate and ecological impact
• The outright attack on labor by capital
• The systemic disregard of consent
• The basis in wholesale theft and plagiarism
• The assault on the very concept of objective truth
• The enrichment of the most cringe billionaire fascist motherfuckers the world has ever seen
THERE IS NO ETHICAL USE OF AI
If you ignore all of that, you are trying to negotiate
what level of harm to others you find acceptable for your personal, selfish benefit. You advance the agenda of these hollow-eyed, joyless money-ghouls who resent art, creativity, and life itself.
THE MACHINE IS DISGUSTING AND MUST BE DESTROYED
@jwz would love to pass these out at my upcoming gathering - http://gelconference.com - are they available for purchase anywhere?
Crawford Tillinghast Appreciation Post
@jwz I don't think I've ever read Lovecraft (not a horror fan), but I looked this up and found
"...stimulates an affected person's pineal gland, thereby allowing them to perceive planes of existence outside the scope of accepted reality."
In 1920. If I didn't know "The Doors of Perception" was 1954, and LSD was first produced in 1938, I'd say he was disguising a story about drug use.
Now I'll have to get a copy and read it. Fortunately, it's out of copyright now, so that should be easy.
@jwz Found it. Standard Ebooks has a collection of out-of-copyright works by Lovecraft. (Naturally, work has filtering which blocks that site for me.)
@jwz I'd rather call it "There is no ethical use for gen-AI". As this one is the problematic part, beside the ones for mass surveillance.
As for pattern recognition in science "AI" has proven to be life saving by recognizing cancer patterns before human eyes can.
People sadly hardly can separate the different tools which are just called "AI" these days while they could have their very own name.
Even pen stabilizers or normal non-generative programming tools are just called AI these days. <_<
been on mastodon for 8 years now and i often get asked what the best part of the journey has been - the sponsorship deals? the free cars and houses? the drugs and blowjobs?
no, it's the one to eight faves i get from the same friends whose posts i also fave whenever i log on
dirt forever baby 
Once again I’m asking you; scribbly or polished?
@heyheymomo Polished, so the viewers can focus more on the fun shapes. I think if it gets too squiggly/scribbly it will turn out as "visual noise" in the worst case 🤔
@heyheymomo for the 3 first panels, love the composition and look of polish, foe the last one though squiggly has more warmth to it (I think it's mostly the table, computer and such looking so straight, but I'm not an artist really so I wouldn't know)
@heyheymomo I really like both, but I prefer scribbly because it feels warmer and more full of life.
@heyheymomo It concerns me that although I can see they are different, it took me quite some time to see which one is scribbly. I see it now.
@heyheymomo I don't care which you choose. If you prefer to do one (or the other) so less people will say your work is done by AI, then that's ok with me.
I prefer:
| Scribbly!: | 298 |
| Polished!: | 122 |
Closed
@heyheymomo
Scribbly feels more soft and huggable. Forg's door, in particular, feels kind of cold in the polished one. All straight lines and hard angles.
@heyheymomo I like the composition of the polished stuff more, mostly, but I like the squiggles themselves more than the clean lines.
@heyheymomo I think the scribbly one here is better. It has a human touch. The polished style feels more mechanical. There's something about the more open artistic choices made in the scribbly one that I prefer as well (tree density, layout of bottom left panel, desk of bottom right).
@heyheymomo Whichever works best for you. Really. Really really really. They don't look identical to me but I had to stop and think about it. Imagine your life with one or the other, assuming it's only your efforts that change and not how they're received. Which do you like more?
Also:
You can just flip-flop based on how you feel each time!
Also also:
Which one makes you feel the most like how your comics feel?
Also x3:
What would forg do?
TLDR: The one I prefer is the one that you like.
@heyheymomo I like the scribbly lines! But in the polished version I like the offset tree branches in panel one, and the little bush beside the house in panel two.
Dick Bruna, the maker of Miffy, said:
“Het gaat om een bepaalde bibber (…) Een goede lijn heeft een goede bibber. Maar die goede bibber vind ik wel het allermoeilijkste. In die bibber zit de tekenaar verborgen.”
‘It's about a certain wobble (...) A good line has a good wobble. But that good wobble is the hardest part for me. The artist is hidden in that wobble.’
Oooh, I just realised, because I'm moving off MacOS, I can use themes again!
Here are some random ones from xfce-look
(I recognise that while it is possible to use themes, not many of them are polished, but I really should be able to use Comic Sans as my UI font if I wish so)
@nina_kali_nina Thanks for posting these. My enjoyment of them is vicarious but it's also useful/interesting. Battery ls a bit of a concern for me right now
@jbeiapc I think a laptop that is well-supported should not take a significant hit at the battery life. But I guess I'll try out this and report, too
@nina_kali_nina And use case matters a lot I mostly use my laptops plugged in so normally it's no problem, but when unplugged today realised my recent, barely used, OpenBSD install was hitting the battery at about double the rate than most Linux distros. Most of the time not a problem because I want to be able to use it at full speed. And on a desktop zero problems. Tuning did help it a bit.
@nina_kali_nina On the note of xfce, I assume you've heard of one of the main devs of it being an AI ghoul?
@shinmera I have, yes; I'm thinking of migrating off to icewm/jwm/openbox for my purposes. I suspect all the big toys are moving in this direction. LxQt allows AI contributions, Gnome says "a little bit of AI is okay" (though some parts of the foundation are taking a strong stance against). KDE is sort of kind of holding up, but there's still no strict policy, and it takes a heavy toll on some really wonderful engineers there.
@nina_kali_nina *me whimpering after having migrated all my XFCE to lqxt these last days*
(I erroneously assumed that minimalist user were about human control of their tool... Damn should have checked.)
@otyugh I know, right? The thread in GitHub goes along the lines of "we will not allow slop, and it must be human contributors, but AI assisted code is okay". Which is quite interesting culturally: I've seen quite a few people saying "we will not allow slop!" at the very same time as they were shipping code generated by "agents" and merged without a human review, let alone human understanding of the codebase
@nina_kali_nina yeah I think I read something along that line from the AI XFCE guy that he has never written slope because he carefully review the generated code before committing it. So it's not slope.
Anyway, I'm pretty left in the dust with my environement pick here. IDK what to install at my next "get a free refurbished computer with linux" regular appointment. Kinda burned my effort on lxqt just now T_T
@otyugh I used to be a big fan of not using DEs and just running a window manager, but it comes with lots of trade offs, obviously
@nina_kali_nina ...Espacially when trying to seduce "normal people" (who don't want to spend time unlearning their traumatic education of windows, and are in panic&fear of anything that looks different).
At this point my experience with new user is that all >30y old want to have a window clone. Anything that stray is a big big risk to lose em entirely if they're not specially convinced or in a group dynamic.
@nina_kali_nina (I had my spur of elitism as a uni student with startx/openbox using arch btw - the cringe is here, would be funny to get back for different reasons)
@nina_kali_nina @shinmera
I sometimes wonder if I should set up an emulated Amiga system using my old backups and do everything from there again, the underlying system be damned.
@nina_kali_nina About LXQt, I saw the subject you're referring to (and the guy seem to be a real pain), but... I also had a friend who made a tool to scan sloped software and it didn't rise any flag. https://c.changeme.fr.eu.org/shlop/about/
My guess is that the "Tsujan" guy is a bit of a contrarian prick and has not even had one AI contribution but still wanted to boast about how it'd welcome it.
...Still not good software to invest for the future /sadpony (i liked a lot of UIs there)
@nina_kali_nina the first one looks more like GNOME 2 🤔
@nina_kali_nina From which MacBook(?) to which Linux-based laptop do you switch?
@theaitch from a used MacBook Air M1 that I got very cheap to ... TBD, most likely a ThinkPad X1 Carbon.
Does anyone still use #Snac software?
@the_coffee_lover yes! I
set up a #snac instance just a few weeks ago.
It’s been purring along nicely, without ever giving me any trouble.
all these "AI escapes & downloads confidential data" and it's still not regarded as malware?
@patterfloof It's a chronological thing. AI you pay for in advance, malware you pay for after the fact...
This morning my 4yo went to daycare wearing his spiderman pyjamas and, on top them, his favourite orange 'spinning dress' and some pants that I've mended with a colourful love-heart patch on the knee (at his request).
He's getting to the age where other kids tell him he shouldn't be wearing dresses, or make it awkward.
He asked me "are boys allowed to wear dresses?" and I told him, "well, some people say boys shouldn't wear dresses, but that's ridiculous! Do you know people used to say that girls shouldn't wear pants? How silly! Anyone can wear anything!"
He's asked me to tell him that several times again
I think a lot about bell hooks writing that the first act of violence society demands of men is to cut out their own softness:
"The first act of violence that patriarchy demands of males is not violence toward women. Instead patriarchy demands of all males that they engage in acts of psychic self-mutilation, that they kill off the emotional parts of themselves. If an individual is not successful in emotionally crippling himself, he can count on patriarchal men to enact rituals of power that will assault his self-esteem. "
Kinda shocked how hard it is to find a decent pepper mill without plastic components. Peugeot I thought you were supposed to be the good brand
@christianselig I am still using the one Alton Brown recommended 20 years ago. Plastic but metal on the grinder. Still works great
Mannkitchen. Cannot recommend enough!
@blokey Tempting if not for the price, and loses a few points for still having a tiny bit of plastic inside
@max Have you actually looked inside? Most people I've seen say that until they disassemble it and the internal hidden parts are plastic https://www.commentreparer.com/73330/Appareils-ustensiles-de-cuisine/demontage-poivrier-PEUGEOT?utm_source=chatgpt.com
@christianselig My two Peugeot (one salt and one pepper) are wood & metal, no plastic whatsoever.
@macfixer Have you actually taken it apart? Virtually every one I've seen nowadays disassembled has plastic components or chambers https://www.commentreparer.com/73330/Appareils-ustensiles-de-cuisine/demontage-poivrier-PEUGEOT?utm_source=chatgpt.com
@christianselig I mean, I don't doubt your story, but then again, I haven't needed to take mine apart…
🧑🍳
@macfixer So to be clear then you can't definitively say there's no plastic whatsoever? :p
@christianselig Sigh… the parts that I can *see* are metal or wood 😜
@macfixer Hey if that is good enough for you all the power to you, I'm far more neurotic haha
@christianselig like the grinder itself being not plastic?
@via I don't want any plastic anywhere ideally. Here's a story where despite having a metal grinder, an internal piece cracked off a Peugeot and got ground into their dinner https://www.commentreparer.com/73330/Appareils-ustensiles-de-cuisine/demontage-poivrier-PEUGEOT?utm_source=chatgpt.com
@christianselig thanks now I’m paranoid
@christianselig https://mannkitchen.com/products/the-original-pepper-cannon-pepper-mill do this, I have one and I have yes added my
dewalt to the top to make a FUCK ton of pepper for dishes on thanksgiving.
@bosh Still has plastic inside :(
@christianselig i only saw some plastic up on the inside of the lid, I have a metal housing, metal gears, metal ball bearings, no where the peppercorns come in contact. did they change it?!?!
@bosh See here, where the pepper sits against the burrs there's a plastic ridge, and given that's the area that experiences all the grinding action (the corns being fed into the burr) that's not super pleasing haha https://youtu.be/z20AhzhbK0M?t=181
@christianselig i just took a pic of mine and i had to use one of my olights on the burn my retinas setting to see down in there, i’ve had mine for about 4 years and it gets abused daily. i’ve burned out multiples before this and one i snapped the screwball off the top of it. and this looks great. I really think there isn’t another one on the market with 0 plastic at this point, and this at least for me has been the only way to get the volume and the speed of fresh cracked rainbow peppercorns for my dishes.
@bosh I just don’t get why they need to, there are plenty of hand made ones on the market that are even cheaper (Piper pepper mill for instance) that use zero plastic (and surely for all the years before plastic existed they figured out how to make pepper mills), and that way you never even have to think about the plastic internals
@christianselig i think its like any project, if they didnt use incompatible metals that are not subject to galvanic corrosion, then they were making tradeoffs of weights of other metals, paint options on other metals, costs of other materials, and ending up at the least worst compromises. I am sure you have even seen this type of project paths in your own work. I would say, form when the pepper cannon came on the market, it was just peugot and all the cheap stuff. so I love mine and have only one early production issue with their cap that they replaced with no issues. I am glad there is competition and if you get the piper let me know what you think, as there is always room for improvement. But yet with having mine I can see why alton brown went straight to the pepper drill over a mill.
@christianselig I’m really not trying to be cheeky but a pestle and mortar might the best option. I have the Peugeot that has metal burrs but does have internal rubber/plastic components for mechanisms. Same as my coffee grinder. But I’m saving this thread in case something comes up.
@christianselig Me trying to find an induction-compatible kettle. Everywhere I look there’s plastic in everything. At this point I must plan on visiting local thrift/antique stores since they used to make them nice.
@christianselig Check these out - I have a model from 20 years ago when they were owned by Vic Firth (the drum stick company). They're still made in Maine. It's basically a perfect mill.
All wood and metal from what I can tell (for pepper - the salt one is nylon inside) https://fletchersmill.com/collections/federal-mills.
@stoneacheck Listed as having plastic inner fittings unfortunately :( https://www.bostongeneralstore.com/products/federal-pepper-mill
@christianselig I found the patent. There is one plastic part: a 3-spoke polypropylene/polyethylene support ring that centers and retains the metal shaft. Peppercorns pass through its large openings, but it doesn’t grind them, and the design specifically prevents the metal retaining tabs from rubbing against it during use. Not technically plastic-free, but there should be basically no plastic contact beyond peppercorns passing by it. I can take a pic of mine later?
@stoneacheck That’s reassuring but you never know when the plastic weakens and could crack over time. Someone posted their Peugeot did that and ground the plastic into their dinner, so I’d ideally like to just avoid plastic parts period if that makes sense :/
@christianselig Sorry, missed the last reply. Like I said, mine's been going strong for 20 years (not exaggerating, I got it in 2006) and you can see it has basically no wear and tear. There's always a mortar and pestle if you wanna go old school and have full control.
@stoneacheck that wood is beautiful. I ended up picking up a Piper since apparently no plastic whatsoever, I’ll try to do a review on it!
@christianselig
It is not cheap but we have been using this on for 25 years. Perfex from France.
https://www.francecorner.com/2625-perfex-pepper-mill-2-sizes.html#/608-dimension-large_model
@christianselig Certainly pricey, but I can personally attest to its crazy high build quality: https://crustmill.com
@christianselig The most annoying thing is that metal burrs are not a whole lot better as you slowly grind metals with the pepper,.. we did some research a while back and our conclusion was that the "old fashioned" way with a mortar was the healthiest option.
@christianselig pepper cannon by Mann kitchen
@christianselig There are a few pepper mills that are basically repurposed coffee hand grinders and they work amazingly well.
poll: when you create an SSH key, do you set a passphrase?
| yes: | 583 |
| no: | 640 |
| it depends/other: | 577 |
| what?: | 44 |
@b0rk all my SSH keys with full access to the target account, used interactively, have passphrases. But when I need something automated like a cron job to SSH from one machine to another, it needs a passphraseless key, so instead I limit the capabilities of that key by only entering it in authorized_keys files with lots of restrictions, including a forced command.
My aim is that every key _either_ has a passphrase, _or_ very limited usefulness.
@simontatham interesting, i didn't know you could put restrictions in an authorized_keys file!
@b0rk the typical thing I write in authorized_keys looks something like
command="/home/simon/adm/some_script",no-pty,no-agent-forwarding ssh-ed25519 AAAAC3Nza[...]Lc7k key-comment
and the script it points to is designed to do only the minimum thing that I need the cron job to be able to do.
@simontatham @b0rk Yeah, very often I'll use rsync in full verbose mode to work out what the command is that belongs in there for an automatic sync job in cron.
@spacehobo @b0rk for rsync in particular, there's a wrapper script 'rrsync' designed to make that easier. (But I think I did the same thing before finding that out!)
@simontatham @spacehobo @b0rk I came here to point out rrsync, which is extremely handy for this kind of case.
@simontatham @b0rk If you can (OpenSSH >= 7.2, but that's ten years old and most systems should have it now), I recommend using the `restrict` option rather than applying individual restrictions using things like `no-pty,no-agent-forwarding`. You can always write `restrict,X11-forwarding` or whatever if you really meant to enable a specific capability rather than just having forgotten to disable it, but this is usually shorter and safer.
@b0rk @simontatham
In practice, I rarely ever use passphraseless/restricted because I just use an agent for everything. Only occasionally do I need a script that is completely automated, usually for file transfers.
Co-worker: I completed the project ahead of schedule with this simple hack
Me: I told you, I’d rather you didn’t refer to me that way
It is possible to have community-oriented software projects that don't feed the LLMs?
#FOSS #LLMs #OpenSource #ClosedSource #commons
| no, we have to accept all our code will be stolen: | 7 |
| maybe / yes (I will describe how in a reply): | 9 |
Closed
I'm open to any solution, everything from data poisoning, to closed source community models, private networks, alternative programming paradigms, guerilla warfare against data centers, fundamentally incompatible computing paradigms, ham radio, astral projection, summoning ancient deities, the overthrow of capitalism, ???
Having all you need (spare tire, jack, etc.) are you able to change a flat tire?
Pls boost for wider sample, thx!
~~
Avendo tutto quel che serve (ruota di scorta, cric, ecc.) sapete cambiare una gomma a terra?
Per favore, boostate per un campione più significativo, grazie!
| Yes — I'm > 59: | 77 |
| Yes — I'm <= 59: | 248 |
| No — I'm > 59: | 6 |
| No — I'm <= 59: | 64 |
@GustavinoBevilacqua Been a while since I've needed to on a car, but probably every other month or so on a bike.
I read the noun phrase "AI supply chain," and I am now damaged at an intellectual level.
@xgranade i'm already zoning out remembering the corrugated steel ceilings in the "offices" where people in west african countries go to look at horrifying things to train automated moderation systems and thinking, "oh, would you really like to discuss at length the AI supply chain?"
(unless they simply mean the compute. in which case, the same phrase applies,)