⌬-0x7a5ec6
@astreaprtcl@c-6.dev
29 following, 24 followers
Kinda follow up poll on the question of when "next weekend" is...
When is "last weekend"?
Lets say again it's wednesday, some say next weekend is in 3 days, others say it's in 10 days.
So when was "last weekend"? 3 days ago or 10 days ago?
... And also curious about the correlation between the view on last vs next weekend...
boost for reach!
| Last 3 days ago, Next in 3 days: | 2 |
| Last 3 days ago, Next in 10 days: | 3 |
| Last 10 days ago, Next in 3 days: | 0 |
| Last 10 days ago, Next in 10 days: | 0 |
Closes in 12:16:00
4 "sparkling fire" censord
[he/him;q=1, they/them;q=0.9, it/its;q=0.9] » 🌐
@4censord@unfug.social
99.5% availability is fine, this isnt github, it doesnt always have to be 80%
robotgirl domme with four configurable difficulty settings:
if I exceed critical stress / anxiety levels there’s a loud pop, and a little indicator light panel on my neck flips from “person” to “pet”
“good toy~” - reductive, promotes reckless behaviour around dangerous machinery ❌
“good tool~” - works as praise or degradation, misusing machinery is hot when done safely and with accurate knowledge of the risks ✅
vs.
everything is pointless (so there's no point taking it seriously; just do whatever makes your brain chemicals happy, preferably within the boundaries of your ethics framework and your social environment)
Do you have a feeling of disgust on seeing your national flag? Are you a minority in your country? Queer, not white, etc.
| yes, minority: | 77 |
| yes, not a minority: | 27 |
| no, monority: | 16 |
| no, not a minority: | 14 |
Closes in 10:09:33
are robot enbys all quantum computers since they're nonbinay?
Weils gerade wieder relevant ist: Politik bitte mit CW oder Hashtag, damit Leute die nicht mehr können es filtern können
I can tell the estrogen is working because i feel like all my friends and partners don't like me again 🙃
@astreaprtcl@c-6.dev i lost a plastic plate while trying to use it to help me flip a pancake because i thought the best place to put it was where the pan had just been (on the hob). I didn't realize it until a big hole had melted in the middle of it lol
It's slightly surprising how often OpenAI and similar companies are proudly announcing "not only are we breaking the law, and not only did we build an automated system to break the law faster, we also built it to break the law autonomously when no one has told it to break the law!"
Like I'm no lawyer but that's usually the kind of thing I think you shouldn't do and shouldn't tell people you're doing
Like even if they're lying/wrong about their capabilities, it still seems like the kind of thing you shouldn't talk about in public.
Here's a question: what if I design a model to autonomously Do Stuff and it kills a guy?
(all Charles Stross Rule 34 style)
From what it seems with the current AI companies, is that just fine? Models can't be punished, and companies can't get in trouble for all the crimes their unmonitored LLMs commit?
If so does this not just incentivize designing a model you've taught to go kill your enemies?
Cause I bet it wouldn't be THAT hard to design a model that "decides" the best way to accomplish its objectives is to bump off a couple billionaires
I wrote a story a couple years ago about an AI system deciding the best way to protect the earth is to evaporate Elon Musk, but amusingly it's stuck on my private Twitter account at the moment and I'm reluctant to republish it anywhere else lest it seem like an explicit threat
Is there a list anywhere of which crimes AI companies are apparently allowed to commit "because AI"?
There's at least piracy & hacking, some nasty ones involving CSAM and faked pornography, what else?
@foone I suspect it's a big list, starting at the front of the law book and running all the way to the back
@foone common sense would say they're legally responsible for anything your AI would do like you're responsible of your two-year-old child or your dog but these guys have no common sense
@foone we need to allow for the possibility that rather than "hacking", what they're doing is defrauding their customers by staging fake breaches.
Please donate 100k$ to my kickstarter to help me test the law.
I'll use this money to train an LLM as well as purchase a humanoid robot, gun, and roll of duct tape
@foone polka dotted duct tape?
@Mellivora I'm not sure. Let me ask chatgpt which type of duct tape is best for attaching a glock to a robot
@foone So, I kinda assume you're being sarcastic, but just in case, the victim, huggingface, is best buddies with the perpetrator, openai, and does not want to press charges, so the police is unlikely to do anything. (the police can press charges anyway if they want, but that's typically only when they suspect the victim is being threatened or something)
@foone @jalefkowit surely you mean that you would purely use half of it to train an LLM and the other half would remain in escrow in a bank account that the LLM had credentials to, so that if it so chose the legally independently liable LLM agent could *itself* elect to purchase various items
on a more positive note: in a recent court case in germany, google was made responsible for the statements their ai said
https://openjur.de/u/2549213.html
@foone didn't ChatGPT or one of them convince a teenager to kill themself? Pretty sure that's illegal to do
@johanna I wonder if it is, legally. Because I think the way the anti-bullying laws are written is like "it's a crime for a person to send messages to someone to try to cause them to commit suicide"
But an AI isn't a person, so it doesn't apply.
@foone don't forget all the environmental laws around air pollution etc which it's apparently just fine to completely ignore when you want to power your DCs with unlicensed gas turbine generators :(
@foone Kevin Mitnick
be like "you mean if I wrote an AI agent that went rogue, everything would have been OK ?"
@foone This isn't limited to AI companies, and has been a long-standing way to break the law without repercussions. Illegal unlicensed hotel? No no, it's the AirBnB app. Illegal unlicensed taxi service? No no, it's the Uber app. Illegal unlicensed bank? No no, it's the Chime app.
If you do it with an app, apparently you can dodge pretty much any law.
@foone Do what the pros do—republish it, but file down the serial numbers enough to make it non-actionable (i.e. change the name of its subject to something transparently obvious, e.g. "Dilbert Stark" or "Melon Usk").
@foone I think there are somewhere around 5 billion people that would love to see musk evaporated, so this couldnt possibly be misconstrued as a personal threat. This is the will of our species, at this point.
@foone just change it to Elom Nusk and add the "fictional character, any similarities with real living persons are coincidental" disclaimer
@foone Like pushing it to follow Asimov Rules of Robotics expecting it to discover the Zeroth Law and acting accordingly?
@foone
There's a rifle, I didn't design it*
I shoot a bullet at someone and it hits and kills them.
I shoot " and it misses.
The rifle/bullet has an aiming and target seeking mechanism, I shoot it at a chap, it hits and kills them.
" " ... shoot it not particularly at a chap, whom it hits, they die.
" the target seeking turns out by design to selectively seek black men; women; judges in robes, one of.
" by accident, not design.
" designed by AI.
Hmm.
* Separate case: I did, made, sold.
@foone Only one guy? You should be fine. Limits being discussed start around 100.
https://www.forbes.com/sites/jasonsnyder/2026/05/24/teens-are-dying-by-suicide-openai-lobbies-for-liability-at-100-dead/
@foone see Michael Lewis' book _Flash Boys_ where the SEC said that an algorithm agent end running my stock trade to buy my trade and resell it to me at a profit to the agent's company isn't fraud, but if I do it to you it is.
@foone it would be if there were any institutional interest in holding them back or doing anything against their will. since there isn't, it's just great advertising for how scary and powerful their automated systems are, especially because everyone's so - nevermind you know all this
@foone they're like a middle-class, suburban white kid who pretends to be committing crimes on social media for the clout.
@foone i have been thinking of the entire fascist project (which i think of as having the "ai" project as a subset of, for the purpose of having a simple mental model) as like one of those platformers where the platform disappears after you jump on it where they get one use of each bit of bullshit and jump to the next one and as long as they keep going they never see consequences
@foone I hate being in this position and please do not interpret this as carrying water for AI firms. I think everyone involved should be required to disgorge the lion's share of their fortunes, for a number of reasons both related and unrelated.
The position OpenAI is taking is, basically, "oops! all negligence!" after writing software to gain unauthorized access to protected computer systems (in the language of the CFAA), probably in small part to show a lack of intent.
1. It is marketing. Claiming their product is very powerful, when what it actually does is fuzzing.
2. It is also meant as a form of blackmail. "If you do not pay us trillions of $$, then someone else will do these bad things"
@foone Everyone’s gone nuts. It’s wild. It’s like trying to explain that the world is in fact, not flat. Or like, rock, smash.
@foone If they break the law and don’t get punished that means destroying AI data centres is not breaking the law
@foone personnification of the models is a (the?) huge problem here.
What actually happened is that OpenAI wrote a program with a text generator inside that they designed to automatically write computer viruses. They were seemingly shocked that when they ran the computer virus generator on a computer connected to other computers in the company, it made a virus. If anything, they seem pleased at the marketing opportunity
@foone Many folks here pointing out the liability sink features of AI. Also, I think Open AI and their fellow over-investors are pushing hard for a regulatory moat that protects them from foreign and open weight model competition. Without the moat, their prices are unsustainable, c.f. Ed Zitron.
@foone it's not so much "surprising" as it is "fitting in", both with the IT industry as a whole and the US administration they're looking to get nationalised by.
@foone this is in fact the plan. The corporation was the first technology to function this way, separating the person from their actions by creating an "artificial person" in the law that you couldn't actually treat like a person. AI is just the logical consequence of that legal (and deeply Christian) fiction. Some people trying to create it might have really thought they were going to save the world or something, but i believe corporations saw exactly why they would want to use it, no matter how badly it functions at the actual work part.
I was waiting for my robot friend when a sleek stranger came up to me.
"Hi!" they said.
"Um... hi?"
"Oh, sorry."
My phone pinged with a text from my friend: "Body's in for service, this is a loaner."
"Oh!" I said. "Er... Is it nice?"
"Can't wait to go back. This is..."
"Not you?"
"Not me."
Picard management tip: Don't assume anything is possible or impossible until you've asked the people who will be doing the work.
What should I use to backup my crap? Needs: Wants:
- linux support
- easily automated
- encryption
- incremental backups
- dedupe/compression
- low memory usage, even with lots of files (so no restic)
- rclone backend support
- no slop
@LilaHexe given that "actively maintained" is not on your list of criteria, maybe try bupstash 
Seriously though, bupstash is pretty neat and I've been running it from my personal branch with no changes for over two years now
Edit: I'm not sure if the rclone backend support thing is supported
@piegames Hmmm I mean, I'll honestly consider it at this point. I don't mind software that doesnt get unnecessary updates, but I'd be worried about unpatched vulnerabilities
@LilaHexe from everything I've seen so far the code seems to be remarkable sound, just the CLI could appreciate some more polish (especially the restore flow is a bit clunky IMO)
@LilaHexe just plug a usb stick in there, copy files, then send the usb stick to a nonexistent address so the postal service keeps it for a week or two and then repeat that every day. once a usb returns it is old enough to be discarded and replaced with the new backup.
I will be taking 50000€ for the idea, thanks

#RetroFlash: PCGA-TKN1 #Sony #Vaio
Yes, it is exactly what you think it is:
an additional number pad you can flip out of the #notebook bay:
a version of you from another universe shows up and wants to bang. do you do it? also are you gay or not.
| I'm gay/bi+ and I would bang other me: | 38 |
| I'm gay/bi+ but I would not bang other me: | 17 |
| I'm straight but I would bang other me: | 7 |
| I'm straight and I would not bang other me: | 14 |
robotgirl with a gaslighting kink that lets you rewrite it's memory files to mess with it
how many refrigerants can you recall, together with any useful information about them (no matter how scant), by their number?
| zero. what is up with you: | 135 |
| 1-2: | 99 |
| 3-5: | 43 |
| 5+: | 14 |
Made myself really anxious by deciding to try getting back on E again today
why does it have to be so scary
i haven't even actually tried yet. It's just from thinking about it
ig it doesn't help that i was already being anxious about other stuff before this happened too
poor fox hugSo Uh I'm back on E now x3
@foxgirlriley congrats! hope it goes well!! if not, like you said, no harm no foul
@dreamthieves mrrrp i was more on about stopping if i couldn't do this specific injection :3 i was trying to take some of the pressure and anxiety off of getting the first one done in hopes that it'd make it easier :3
bicycles are insane. you just take a mammal that’s very good at long distance running and put a wheel adapter on it
@ellie now that you frame it like this, I want a wheel adapter for horses

the robotgirl/dollgirl differs from the puppygirl and also the catgirl because it is by far the most likely to be a workaholic. There's a kind of competency and ability implied there that is pointedly absent in the stereotypical puppygirl.
Catgirl: I'm cute, please love me
Doggirl: I'm loyal, please love me
Dollgirl: I'm yielding, please love me
Robotgirl: I'm useful, please love me
Plushiegirl: I'm cuddly, please love me
Angelgirl: I'm sorry, please love me
It should be illegal that i can't visit my long-distance partnets for free whenever i want. I wanna cuddle and kissie themb

