It's only enforced because of Rust's strict type system. Python, on the other hand, lets you do whatever you want by comparison, and complains only at runtime. I've seen far too many **kwargs for my liking.
And FD types (e.g. OwnedFd, BorrowedFd) are already in std.
My example would be a thin wrapper around these, most likely. It's only an example of what I'm trying to convey, though.
AGI wouldn't be about a model that knows everything about language and every advancement in every field, but rather a model that is better than humans at finding solutions to problems
A LLM (or any other kind of model) that cannot adapt to changes in a field cannot perform better than humans in that field after that field experiences significant changes. Any such model would eventually degrade in output quality over time.
Also, AGI (artificial general intelligence) usually refers to an AI capable of performing all cognitive tasks at least as well as a human. It's as much of a buzzterm as "AI" is, of course, so there's an endless number of definitions for it. Such an AI should be capable of, at minimum, adaptation over time.
The point of a "smarter" model is not that it knows all the facts, that would be wasteful as it is trivial to look up facts at inference time.
An omniscient model would be impossible, but that's not what I was referring to at all. LLMs these days fill their context windows with relevant information through careful prompting, tool calls, and so on. This is generally how a model is supposed to adapt. Context windows are bounded in size, though. It would have an increasing amount of information to include in that window over time, meaning the amount of data it needs to fit in the context window is unbounded.
Unless someone creates a LLM with infinite context (which would require infinite VRAM), such a LLM can never exist. Therefore, a LLM trained today will never be equivalent to (or better than) humans at all cognitive tasks for the entire future of humanity. There will always come a point where such a LLM's output quality degrades, and it can do nothing to resolve that.
Edit: Here's a simple example: a new written language emerges with all the complexities of a language like English. Humans can learn that language and communicate in it. A LLM cannot.
Many of their TOCTOU issues are something a type system can help with. Require operations to execute on a fd handle directly rather than using convenience functions.
rs
let fd = FileDescriptor::new(path);
fd.delete()?;
fd.create(mode)?;
let is_root = fd == FileDescriptor::new("/"); // does (dev, inode) comparison internally
// etc
The uutils devs would need to create that themselves, but OpenOptions seems to get them part of the way there at least.
While I'm interested to see the proof, it's more of a formality. It doesn't take a PhD to ask what happens when the "AGI" LLM is trained on out of date information. They don't learn over time, and they have a limited context buffer. At the very minimum, it would run out of context just keeping up with changes to spoken language over 30 years, let alone advancements in fields, new fields, and so on.
Nushell - because open blah.json | get foo.bar.2 just works. It also just works with yaml and any other formats I want to support (you can define custom commands to support any extension you want).
uv/fnm - good tools for Python and JS
Starship - nice looking prompt with useful info
zoxide/fzf - because z myproject saves enough time and effort to justify using it over cd most of the time
Carapace as my default completer for better completions on most common tools
Gitui - super nice git tool. I still use git directly a lot, but Gitui's interface is more convenient for staging changes
Patel hasn't made any statements towards being a Nazi. Fuck Omarchy, and fuck him for supporting the development of it, but that itself doesn't make him a Nazi. Framework still makes a damn good laptop, and it has forcefully pushed laptop development in a positive direction. Lenovo sort of did too with Thinkpads, but that's the closest equivalent that comes to mind, and even that's a stretch. So, unless there's someone better out there, I'm still recommending Frameworks to people.
WOTC has a lot of very passionate devs working there, and having looked at job postings in the past, they don't receive anywhere near the recognition they deserve. Here's hoping this gives them the chance to be properly compensated for their work! It's still a miracle they are able to build and maintain a rules engine for MTGA that works as well as it does.
Last time I tried scanning and printing on Windows, it took me over an hour to get the device recognized, the right drivers installed, the printer to actually receive the print job, and so on. Printers are just shitty pieces of hardware, Linux or not.
What you're referencing is distillation. Anthropic even has an article on distillation "attacks" (as if they have some divine right to the data behind their models) that goes over it a bit.
This is the bullshit they're telling themselves behind closed doors
Birth control pills usually come with a ton of risks. If you ever actually read the box, there are so many things that can go wrong with it, some of which I've seen firsthand. That's not to say those things will happen though, and like all medications, the obvious solution is to just stop taking it if you experience those side effects.
Using that to say contraceptives are bad is stupid, though. It's only one of many types of contraceptives. Also, birth control pills serve an important role for many people. Even ignoring their importance as contraceptives for the huge population of women who do not experience those negative side effects, they're also given to stop periods to help with stuff like endometriosis, where having a period could result in internal scarring.
All this to say their bullshit is small truths mixed with larger lies. For example, what they're saying about cancer is bullshit, but blood clots might be a real risk.
Are any of the reported polling numbers reliable? I'm not too sure who VTsIOM is, but if they're state-owned, then I'm inclined to believe that anything they say is bullshit.
How do you recover an account on the other providers? Do you have to provide the same recovery email you set before during account recovery? If you hash the email, you have no way of reading it anymore, so someone has to provide it to you again.
I've had laptops where this is a BIOS setting. A simple switch on the side of the keyboard to flip which keys are on the fn layer solves the problem too. Same with key remapping, for keyboards that support that (and really any nice keyboard should or it's not worth the cost).
Also,i'm sure you know this, but security through obscurity is a poor systems design choice in almost all scenarios.
The only time I can think of from the top of my head where obscurity aids security is when secret keys are kept obscure. This isn't even what people mean by "security through obscurity" though, so I'd actually beg someone to give an example where obscurity is actually beneficial to security and doesn't just give a false sense of security instead.
That's not to say everything can or should be open source, of course, just that relying on it being closed source for your application to be secure is a good way to open yourself up to attacks.
What does any of this have to do with GPL or open source licenses? Military applications all have strict validation requirements that rule out the majority of open source anyway, and your first example doesn't even explain how the software being open source would be dangerous at all. Actually, for that matter, nor does the military example. Encryption doesn't work because the other party doesn't know your algorithm lol, it works because the other party doesn't know your secret keys.
If you're referring to GPL variants, that depends. You can absolutely use GPL software and libraries with closed source software. You just need to separate the GPL portions from the closed source portions with some sort of boundary, like running it as a service of some sort or turning it into a CLI tool. You're just not allowed to create derivative works of GPL software that isn't also GPL.
Also, there should be nothing dangerous about open sourcing code (unless you're referring to financial risk to the business I guess). Secrets should never live in code, and obscurity is never secure.
It's only enforced because of Rust's strict type system. Python, on the other hand, lets you do whatever you want by comparison, and complains only at runtime. I've seen far too many
**kwargsfor my liking.My example would be a thin wrapper around these, most likely. It's only an example of what I'm trying to convey, though.