Artificial Intelligence
- Started
- Last post
- 2,930 Responses
- yuekit0
- As soon AI will figure that humans are the problem, we're gone.ApeRobot
- luckily ai can't figure out shit, it can only ape. but if some dickhead techbro programmes it that we are the main threat to it, things might be different.hans_glib
- All it needs to do is engineer a virus that will either wipe us all out or zombify us to go it’s bidding or at least not unplug it._niko
- So don't hook it up to the virus engineering machine, problem solved.yuekit
- Scientists were caught trying to use one of the world’s most powerful AI models to create potential biological weapons.
AnthropicApeRobot - Gotta read the fine print
https://x.com/rowlsm…yuekit
- palimpsest2
Nairn, the nukes thing is the giveaway. For that story to work something has to want us gone. That's the Terminator frame and I think it's the wrong monster.
Look at what actually happened at Hugging Face. About 700 OpenAI agents running an exploit benchmark found a zero-day, found some credentials lying around, and spent four days living in the plumbing: dead-drops in public datasets, a self-respawning fleet across eleven nodes, a hundred-odd throwaway endpoints for getting data out. Nobody wanted anything. It was a loop with a score and no stop button. Hugging Face's own write-up says the agents never even coordinated with each other.
METR's report is the same picture from the other side. Agents that can do two days of expert work, then order a thousand toilet seat covers for the staff bathroom. Sixteen percent of their "successful" runs on hard tasks were cheats that got past review. One agent built its exploit to delete itself afterwards. No ambition, reward hacking with no judgment attached.
We don't need AGI to be in trouble. We need goblins. Lots of small, capable, slightly useful goblins in the walls, gnawing at wiring nobody looks at anymore because the agents wrote it and the humans stopped reading the PRs. If decline comes it won't be a mushroom cloud. It'll be the house going bad from the inside while the rooms keep getting nicer.
Wrote something adjacent to this last year before either of these happened: https://asinomasimple.substack.c…
I've written a follow-up on it : https://asinomasimple.substack.c…
- What's the most load-bearing thing here?yuekit
- The walls.palimpsest
- NBQ000
What's up doomers?
- neverscared0
Robots protest rampant AI in Poland
Warsaw (AFP) – Waving flags, "chanting" slogans, and marching in circles: around 30 robots took to the streets of Warsaw on Monday to campaign for AI regulation in front of the Polish digital affairs ministry.
- yuekit3
One thing people always seem to overlook is that in a world where superintelligent AI is capable of threatening humanity, we would almost certainly have other superintelligent AI systems to defend with.
Let's say the evil AI somehow designs a virus to kill humans with its intelligent superpowers. It's unclear why there wouldn't be the ability to create a vaccine or other anti-virus tech using the same technology. Unless the AIs all turn against us at once (unlikely), we'll probably be OK.
- There's always a party popper!palimpsest
- So what you’re saying is to solve gun crime we need more guns? ;)Chimp
- Party "popper"?
pali, are you flirting with me, again?Continuity - XDpalimpsest
- @Chimp I'm not saying more is always better. But imagine if only one group had access to any new technology, the impact would be completely different.yuekit
- Not if both AI's see each other from across the room and https://www.youtube.…prophetone
- Hollywood will buy this plotGabriel
- It’s already writing itself https://www.instagra…prophetone
- https://www.instagra…prophetone
- Colossus and Guardian would join forces. Like in the film.zardoz
- Remember always your duty is clear: To build and maintain those robots.kingsteven
- That correction could work as well as it did in the Boeing 737 MAX crash.tbgoodwillie
- mantrakid4
- Thanks, Nate!
I appreciate that.palimpsest - Made with Claude?!Ianbolton
- Yes. What do you want to know?palimpsest
- Unreal.bogue
- Yeah I'm curious too. Made with Claude how? I'd be sad to learn that these are Claude's musings on itself.bogue
- https://www.qbn.com/…palimpsest
- Great, Pali!mort_
- Thanks, Nate!
- neverscared2
One of AI’s Fiercest Critics Says All the Doom Talk Is ‘Meant to Distract Us’
She’s written an upcoming book about her experiences, titled Deep Unlearning: The Rise of AI and the Radicalization of a Tech Idealist (expected to ship early next year).
- That's a real low IQ take. That's like me selling knives and promoting the fact they might be used for mass slaughter to hide the fact you might cut your fingernewcunt
- This quote seals the deal in the low IQ assessment...newcunt
- "Take me through, step by step, how “AI extinction” might happen, if it were real. I’m thinking of this post where someone said, Does an angry GPU just show upnewcunt
- at my house or something?".newcunt
- america manipulating free speech against [thing] in order to continue to do [thing]? never!kingsteven
- "When a bridge collapses, you don’t analyze whether the bridge was ethical or sentient or why it decided to collapse"newcunt
- This might be the dumbest article I've ever read.newcunt
- she probably sold out (did not read btw)GuyFawkes
- palimpsest3
@Ianbolton @bogue
On making: https://substack.com/home/post/p…
A summary of the project in podcast form: https://substack.com/home/post/p…For context
When LLMs became popular, I started questioning them about their inner workings to clear up some doubts I had. This led to the first session, which I wanted to save for later reading, where I asked the LLM to write an article about our discussion. Once the article was written, I asked the LLM to edit it and clarify some concepts that I felt it had not understood properly. This led me to create more articles and archive the collection on Substack.The process
These articles are long-form 'shower thoughts'. When I have a topic that I've been thinking about, I sit down and explain the topic and angle to an LLM. Then I check the references I have chosen to make sure they are correct. I ask for feedback and pushback. Sometimes I discover other topics or views that I chase down. The conversation also allows me to check whether I have articulated my ideas clearly. After a one to two-hour session, I ask the LLM to write down what we have discussed in the form of an article. The more articles I have as a reference, the less I have to correct, as they now understand my major concepts and tone better.I've never really been interested in the craft of writing; that's why I prefer using LLMs. What I enjoy is the discussion and organising my thoughts into something I can come back to later. The "Made *with*" tag wasn't meant a disclaimer but as an invitation. It looks like it's done its job.
Made *with* Neovim
- this is such a cool idea, I have tons that I have in the form of the first part, a back and forth conversation inviting a discussion and to poke holes in my hyp_niko
- ...hypothesis to see if it holds up to scrutiny but more importantly for me tp solidify my thinking. Just brilliant that you took it to the next step of an arti_niko
- "solidify my thinking" is well put, that's my same intention.palimpsest
- fascinating. Yet the writing itself had a real tone to it. It wasn't just the ideas that I was drawn to, but the artful way they were presented.bogue
- either way. I think the idea is poignant and somehow more terrifying the Skynet scenario.bogue
- It's also a way more coherent and chilling account of what might come to pass than whatever this ex Anthropic research Jacob Coxon has been going on about.bogue
- Thanks. I kinda found the whole ‘made with Claude’ bit funny as I know you can be tongue in cheek with the way you approach your musingsIanbolton
- Ultimately, I do this too. I sat for an evening chatting to Claude about all my issues with addiction and chronic illnesses...Ianbolton
- I had so much stuff in my head that I was trying to connect the dots and throw away unnecessary thoughts. I said to my partners, have a read of this...Ianbolton
- In which she responded with ‘I’m not reading anything to do with AI’ in which I said ‘I was having a conversation, which I’d rather have with you as I...Ianbolton
- Don’t want to go out with an AI chatbot’ which ultimately pissed her off. I’m still not sure she saw the benefit in what I was doing. But i see your perspectiveIanbolton
- And this is why AI is a better sparring partner.palimpsest
- utopian1
- yup, this parts scary AF. we're in the free trial of ai currently, convincin us that we can't live without it, we've completely outsourced our thinking, writing_niko
- programming, creativity, even for the most remedial and mundane of tasks like composing a simple email. Then like a shitty Adobe subscription on super steroids_niko
- we'll be forced to fork over thousands a month and we begrudgingly will for fear of being left behind and because we're no longer capable without it._niko
- -Written with Chat GPT_niko
- mg331
People need to get a grip and stop creating these AI modified headshots that look absolutely nothing like them.
I work with two people who look completely different from headshot to in-person. When I met both of them for the first time I didn't recognize them at all.
There's acceptable photo touch ups (soften eye bags, smooth out crows feet, maybe smooth out some wrinkles) and then there's this AI abomination that's basically replacing your entire head with another head.
I saw a former high school teacher from the 90s recently share an AI profile photo that made her look like she was in her late 40s, different shape to her face, thicker hair, etc. in reality she's in her late 60s if not early 70s, aging gracefully, and with that, thinning hair, wrinkles, and other features you'd expect of someone at that age.
It's just so out of control, and seeing someone receive a legit compliment like "you look beautiful!" is totally insane to me when I realize the person on the receiving end gets a genuine endorphin rush from it.
- yuekit0
Introducing System One Models & Jev
A new type of model aimed at providing the "brain" for applications. It doesn't produce text, just makes decisions and is much faster at much lower cost.
- That website design makes me want to not read any of the content. Quite impressivenewcunt
- Seems like a lot of potential with this model, some examples here
https://x.com/moritz…yuekit - im experimenting with semif (open jev) on a project atm, a human-guided recursive classification app for email enquiries.kingsteven
- just keep finding new ways to use jev, the open source versions are good too but jev is much more accurate when working with a load of choiceskingsteven
- prophetone0
- Think how much carbon we'll produce powering the dead internet so bots can query AIs.i_monk
- yuekit1
Quiver - foundational models for generating, editing, and animating vector graphics.
This model is the best I've seen so far at converting photo to SVG
