This is a cool idea, and I love the hacker energy behind it, but I agree with some of the other comments that the receiver would have to trust you quite a bit to allow hardware into their house. Even with the SSH view-only accounr, you could be putting the SSH server in a VM so that the client thinks they're seeing the whole OS but they aren't.
Alternative idea would be to provide an open source ansible script or docker compose file to spin up everything you need, and let them verify and deploy it themselves. Then they give you the wireguard configs to remote in.
AI is really fast at searching forums though, can usually dig up posts better than I can. Doesn't help that AI slop has filled the internet making manual search harder
Something I should point out just in case is that 50% success rate is the same as random chance. So when the study says:
AIs were classified as AI 56% of the time.
That really means that humans could not tell the difference between AI and human, since their success rate was almost the same as random chance.
6 language specific tasks. They were programmed and set up specifically to perform whatever specific one task they had to do at the time, that is what computers do.
Sure, and that task was "imitate humans", and these computers seem remarkably good at it. That's what I'm saying. I don't care that these AI work differently than humans. They can imitate humans so well that humans can't tell the difference.
Among AI models, Blenderbot stood out: in AI-AI conversations, it was judged human 67% of the times - more often than actual human-human conversations. In human-AI conversations, human participants were labeled as humans 68% of the time, and AIs were classified as AI 56% of the time.
That sounds like a lot of work to just read some docs. RAG doesn't require any training at all. I assume there are other approaches as well that would be faster than LoRA
I'm aware of how fedora atomic distros work, and that some things are configurable. My point is that a Bazzite user doesn't need to touch the cli or any config file, to just start using their desktop. I think those of us that have been using Linux for years, forget how rare it is for Windows and Mac users to touch the cli or config files. That's the experience that the Universal Blue ecosystem seeks to provide.
A bunch of containerfiles, scripts, ujust files, ... not hard to understand but kinda messy.
I prefer containerfiles tbh, because I use them for my server too.
And you kinda need a server. On NixOS you do this on your machine.
Technically you could set up the ublue infra on your own machine. Building and serving container images inside a container on your own host. Though it's more overhead than NixOS I'm sure.
do note that there are presets online and people help you set things up.
I'm aware but from what I've heard it's similar to docker compose configs. You can't just paste them in and expect everything to work. You'd have to copy one person's config for their entire desktop, and never deviate. At that point you may as well just use rpm-ostree, which is designed for this sort of workflow.
Aren't they using the legal system in this case? So it is by definition, within a set of established laws. Though to me it's not clear if the judge will let them actually get this data. If this were Apple, Apple would probably push back. Microsoft might just give it away for fun though.
My point was simply refuting the "no government" part of your statement. There are plenty of authoritarian governments that are just as bad on privacy
Not the commenter you're replying to, but for me the ending of "2001: A Space Odyssey" is clearly some abstract artsy metaphor, so it's ok that it has zero scientific basis. Like ending on a dream.
The ending (the falling into the black hole part) of Interstellar is actually part of the plot, it actually impacts the other characters. If you want to sacrifice all logic for some emotional ending, there's plenty of fantasy movies like that, but it's not my cup of tea.
Maybe with their state being fixed in time; LLMs do not change
Not with agents and chain of thought. Agents can run for hours continuously. So sure, an AI agent is not the same as a sentient human on the time scale of a year, but on the time scale of a few hours, perhaps the AI is "sentient".
The first paper you linked is focused on the mechanism and not the output. It creates a definition of "thought" and then talks about how the AI doesn't "think in the feature space". It only addresses chain of thought at the end, and says that the issue is that it constrains AI to think in natural language only (instead of, say, pictures).
But again, why would thinking in pictures define sentience? The paper gives a rock-paper-scissors example and says that the AI thinks about it in a different way than a human. So what? If a human plays rock paper scissors against the AI and the AI's output is indistinguishable from a human's, 99.9% of the time, why is that not sentient?
This is like saying python programmers are not real programmers. If a python programmer can implement the same program in python, who cares what language they use.
I'd love for a better definition of sentience than "it works differently than humans".
Unrelated. For many intents and purposes the states of the US are in fact united. However the examples in OP are very much not democratic.