Wine/proton are not emulators, they run the code directly on linux, but using some fancy tricks to rrimplement windows dlls into linux native code. If someone embeds linux code in the windows executable you have the same privileges of your user and at tgat point you can be infected with a linux virus.
It’s not likely today because windows is much more popular, but once linux becomes popular enough to be an attractive target you will need to be as careful as on windows.
Very interesting! Thanks for the links and the detailed answer!
Let me preface this by saying that I am generally against AI and I hate LLMs being pushed everywhere, but I currently hate facebook and social media even more than AI.
I think that ad blocking might actually be a good use of a model trained to detect ads: AI models are “black boxes” and it would make it difficult for facebook to find out precisely how the detection works and workaround it. Imagine a tiny classifier running locally, whose only job is to look at a post html or resulting rendered pixels and detecting if it’s an ad or not, and then generating the blocking rules.
It would be quite cool, because it would work on any website without an explicit list of ad-blocking rules that somebody needs to maintain.
Ylvis did a funny and catchy song on Stonehenge!
Fun fact, in emacs they added M-x butterfly as a reference to this xkcd.
(for people missing the reference: https://www.gnu.org/fun/jokes/ed-msg.html )
Yes I never trusted it fully, and I would strongly suggest anyone to run it in a VM or at least a container, where it doesn’t have access to your credentials, browser cookies and sensitive files. It’s very irresponsible otherwise.
Example: around 1 year ago the bot made a small coding mistake and created a ‘~’ folder. I asked it to delete it (not in auto mode) and of course it went with a very nice “rm -rf ~” (which of course I denied).
Modern models are much smarter, but I would not trust the auto classifier to stop everything dangerous and make a similar hallucination slip through.
The linked video at the end of the article shows the trail, and it’s for sure an expert level trail, full of dangerous sections. The video says the full trail is only 30km long, but they reccommend 2-3 days, it’s full of steep sections, cliffs at the side, false trails leading nowhere, unstable terrain, bears… They were lucky they actually didn’t get seriously injured.
You can implement an agent loop in any language, and modern languages can do everything that lisp can, so there’s nothing magic in lisp. At the end of the day lisp is just another weakly typed functional/procedural language.
Personally I didn’t find lisp that much elegant due to the loose typechecking and the excessive parentheses, but the macro system allows some very terse and smart (and mind-bending, as you said) solutions, which do feel pretty cool when they work fine.
IMHO it’d be probably better to give your agent access to a python interpreter because it’s much more prevalent in the training data than lisp…
I self-learned lisp last year using the old SICP book, but I do consider myself a fairly expert programmer, and I felt that book was a bit hard at the end of it.
IMHO the key difference of lisp’s eval wrt python’s eval is that it takes in input a parsed expression (a lisp list) and not a string. In lisp though you don’t have to use eval much, because the macros allow you to transform the expressions from your programs directly (take in input parsed code, return a different code structure to be executed). The quote and quasiquote functions are there to make that easy!
I don’t know javascript enough, but in python you’d have to use eval and use some kind of parsing package crazyness if you wanted to implement a similar macro system!
Not to mention the entitlement of saying “But let’s be honest, if you’re reading this you can probably pay for it,” Not everybody has disposable income, you know.
I stopped reading there.
I agree! On gentoo even if you do manage to break your system you’d also know how to fix it (since you basically spent days installing it by hand…).
I ran it on my pc for a few years and the only source of breakage were nvidia drivers (in the past they sucked possibly even more). I remember running it using VESA drivers for like a week until I figured out how to fix it.
It’s really nice how much you can customize it and make it really fit your own use-case. I once ran it in a fully offline raspberry pi for 5-10 years (without updates) and it was rock solid. The hard disk it ran on and the power adapter died, but the sd card was mounted read-only so it lasted the full 10 years without any corruption. I kept a backup of the root filesystem on the sd-card so restoring it to a new disk was super-easy. I also ran the hdd in btrfs raid-1 mode with a itself (two partitions), so that it could self-recover from random corruption due to power losses (I’m not sure it’s a good idea, but it never caused any trouble) It was just running 1 service and that’s it, without any monitor or keyboard connected. I even made a small program with a single-key interface to run admin commands directly without needing a screen or any login (press “b”=reboot, “s” for btrfs scrub, for example). Trying to customize, say, an ubuntu like that would have been a nightmare. It took forever to compile everything on the raspberry pi though.
If you are into exploring programming languages or writing interpreters, lisp is quite an interesting language. I can also see it working quite well if you need to embed a small scripting language somewhere, because it’s so minimal and easy to pick up, but also very powerful if you go into the metaprogramming side of it.
I don’t see many reasons to use it for anything else though.
The “outside any juristiction” part is probably the main reason here.
Try to seize any evidence from servers that are traveling in space 10000 miles/hour. For now also space weapons are mostly in experimental phase, and I assume would cause a lot of debris in orbit. It makes the thousands of satellites in orbit almost untouchable by any government on earth.
The epstein-class has much to benefit from this.
It might also be playing into the distopyan fantasy of putting a superintelligent AI in earth orbit, so that it becomes unstoppable from earth.
In any case nothing good comes out of this.
They are a big company… Quarterly profits are the only thing they care about. If you trust any big company to forego profits to support a cause you will be very disappointed.
My personal POV is that I’ll welcome the pinkwashed messages rather than having no support at all (something is better than nothing after all).
And btw, I’m not here excusing them for all the other terrible stuff they did. They should be blamed for all the anticompetitive behavior, military involvement etc. I just wanted to say that the hollow pinkwashing is the norm, not the exception.
The trick for windows is to rename the file and kill the process that’s using it / reboot (if it’s some unkillable service). The service won’t then be able to find the file after reatarting and you can delete the renamed copy.
On linux this whole thing is done automatically by the OS: if a file is open and you delete it, it will disappear from the folder but its data won’t actually be deleted from the disk until it’s closed by every program. Some programs even use this trick to create invisible temporary files to offload some data from RAM to disk!
One word: Terraform
deleted by creator
I don’t think I ever worked in a company doing real agile (involving customers early and directly, shipping frequently, gathering feedback, and changing the process if it becomes an obstruction). It always ends up being all about random metrics, velocity reports to management, and “”“requirements”“” dictated from high up. It’s always just chaos that names itself “agile” as a poor excuse for the lack of planning. You get the worst parts of waterfall (lack of feedback and validation) with the worst part of agile (lack of specs and medium-term plan, no project management). No estimations, no requirements, no plan, no coordination between teams, but the deadline is fixed anyway and the feature has been sold to a customer already. The customer is going to see the product for the first time after 6 months of development, after which the priorities shift immediately and no feedback is ever addressed. AI is not going to magically make management write good requirements and have a good planning session with everybody involved… it’s just going to create even more unreviewed AI slop documents that are just noise and will be ignored by most anyway.
I should publish a blog post titled “Using Antropic’s Claude is a Perversion of Writing”. If you are taking AI output and thinking it’s good and engaging, you are making a load-bearing mistake.
Joking aside, the criticism can come only after they implement it and we can see if it’s having any effect on the output quality. I doubt it’s going to be detectably worse.
If anybody is feeling so strongly against watermarking even before we can evaluate how it affects the output, I am only thinking that they have ulterior motives or they don’t want to be caught spreading copy-pasted slop.