Autistic coders are only slightly more rare that autistic rail fans. There's no shortage of you guys in the field.
what you are doing should have a clear connection with some goal, or be clearly a goal in itself, otherwise you’ll achieve nothing.
"Understanding the whole stack" would, necessarily, be the second kind of goal. And you'll never get there as the field is evolving under your feet.
Yes, then you have another problem - how do you choose what you want to do if you don’t yet understand the whole in any approximation.
How did you decide to write English using the Latin alphabet? You did not, I presume, study the whole ancestry of the alphabet back to Hieroglyphs to understand it in it's entirety (did you know that 'A' is an upside-down ox head?), nor did you study alternative spellings, nor did you study linguistics to make sure that English, Modern English in particular, truly, is the best choice of language.
You were able to ignore all that, why are you not able to ignore things elsewhere?
And selective ignorance, btw, is a key skill to aquire as a coder. Encapsulation, abstraction, action at a distance being the root of all evil, all those are key principles to understand and skills to acquire. Why? Because you're not as smart as you wish you were. Being good at ignoring things, being good at saying "if I build it like this, I can from now on ignore the details" is the only way to do anything of any complexity. I don't care how my pants are constructed, about the lubrication the loom uses, I care whether they fit, are comfortable, durable.
When figuring out what to pack for vacation, do you already tetris your shirts and pants? Nah, that comes later. Right now, worry about not forgetting your sunglasses, don't worry, they'll fit somehow.
There’s a barrier a person has to grind through with their teeth before they understand that they want to learn Haskell and what that is.
Nah. Just start somewhere. If you later on realise that your interests lie elsewhere, then switch, but don't fret: If it was interesting enough to look at, how could it have been a waste of time.
How the hell do you cross the gap between these and actual understanding? Other than the blind way of going up level after level, starting with a bipolar transistor, which doesn’t seem easy at all.
You don't pretend that Haskell has anything to do with electrical engineering and then you're golden. You do not need to understand the one to understand the other. You do not need to understand quantum mechanics to understand a transistor, either -- I mean, sure, if you intend to develop process nodes then you better understand quantum mechanics, but if you plan on soldering transistors together until you get an FM receiver? Who cares. Learn to read datasheets, that's the actual skill you'll need.
You choose some random interest and learn it and don't look higher up or deeper down the stack, you respect the abstraction boundaries, until and unless you actually have a good reason to cross them.
The Rust book is also good at teaching coding, but it would be more of a jump into cold water. It's still more about teaching coders about Rust, only teaching coding incidentally to pick up people coming from a variety of languages, instead of getting into the core of computation itself.
Have you already heard of our Lord and Saviour, the Wizard, and the scripture known as The Wizard Book? By the end of it you will be able to write a compiler, be a smith who can forge their own tongs.
(The software to use with the book is nowadays called racket, use #lang sicp to enable the right dialect)
Cause when it stops being glamorous magic, it starts becoming optimized and normal, like painting fences.
The portrait kind of painter working like the fence kind is not a good thing. It's not even a thing at all, it's an illusion.
Also when it comes to accessibility for the aspiring hobbyist coding is very accessible. If you have nothing to start with, sure, more expensive than knitting, but probably not more expensive than acquiring a merino wool habit, and if you already have any kind of computer, any, as well as an internet connection, it's literally free. The tools, the knowledge, everything. Willingness to learn not included.
The question is rather "how many people have a metro station within walking/biking distance" and "how many long-haul trips do you need to make".
Over here we don't set aside half a day (or more) to to drive to walmart to buy groceries for a fortnight, we pick stuff up as we need it when we're out, anyway. Dropping into the supermarket to grab some things is like a five minute detour if you know what you need and where it is. You can spend the metro ride thinking about what to cook, buy what you need, then get going.
According to statistics commute times in Europe are actually slightly longer than in the US, but that doesn't take into account that combining trips is much easier over here and that riding public transport gives you time to, whatnot, knit, biking or walking counts as exercise, while driving a car counts as, at best, nothing, at worst, the road rage will ruin your day.
I'm not saying that you, personally, can flip a switch and make it work for you, on the contrary: The reason that you're not doing it organically is because the infrastructure where you live is right-out designed to not make it work for you. What I suggest is that instead of saying stuff like "It cannot be the case that Europeans are living better lives, they must be imagining things" you say, to your compatriots, "How are those bloody europoors better at this we are supposed to be the best let's figure out how to beat them". Or at least that's how I imagine motivating Americans looks like.
but that’s just not feasible for real working class Americans in the economic system as it is currently
Nothing to do with economics, everything to do with city planning and resource allocation. Public transit and bikes are a bad option in the US because the transit is completely underfunded, "only poor people take the bus", and bike paths, even pedestrian paths (if they even exist) are sent on detours around car infrastructure instead of cutting through everything.
And then you have to juggle picking up your child from childcare, etc with is ridiculous without a car.
My mum did just fine first coming by with the bike, putting me on the back seat, then swinging by the supermarket, groceries in the front basket, later on coming by with the bike, me riding along on my own, still swinging by the supermarket. We were driving on calm backstreets and through a park which was actually the most direct route, much more direct than with a car as you'd have to get onto the collector, first. Got more than one kid to wrangle? Put them in a trailer, or get a suitable cargo bike. They can even have seatbelts.
No, you don't need a warehouse full of washing machines in every neighbourhood. People don't shop for washing machines daily. People don't need cars to shop for them, either, delivering bulky stuff makes a ton of sense. Groceries? Wherever you were that day, a supermarket should only be like a two or three minutes detour.
And it's not like European cities didn't go down the car-centric route, mind you. Difference being we realised it's a stupid idea.
Permanently Deleted
Already there. The blackwall is AI-powered and Markov chains are most definitely an AI technique.
We already can detect direction of infrared radiation, it's called being warm on one side but not the other. Technically also possible by, say, lying half-way under a blanket and half-way not, but sensory integration takes care of the ambiguity.
More interestingly, did you know we can see the polarisation of light?
As a small aside “Open Source Free Trials?” If it’s open source, can’t they just disable the trial part?
Yes. There's a number of projects which distribute binaries which aren't as liberally licensed as the source they're built from. E.g. Ardour is another one. There's a demo version, subscriptions start as low as $1/month, $45 buys you the current major version and the next major version with all its updates, perpetual license. There's also the implicit understanding that if you don't pay up and want support, your bug reports better be developer-grade.
Basically it's a way to get artists who are used to either freeware or commercial offerings to donate. Also as far as DAWs go it's a fucking steal.
Under Article 80, the GDPR foresees that non-profit organizations can take action or represent users.
Emphasis mine. Just in case anyone is still denying that Euro-English is a thing. It provides for it, which implies it anticipats it, so if we get rid of unnecessary Romance it foresees it. Plans for, as Wikipedia puts it, but not only the French etymology (prevoir) works, there's also German vorhersehen/vorsehen which collapse when you calque them into English. Gives Brits all kinds of headaches, they were writing lots of huffy memos before Brexit.
I think there’s some „reasonable” keyword in the right to be forgotten.
The original case was a Spanish cook being haunted by the first google result for his name being an article in a local newspaper about his restaurant going bankrupt decades ago. No scandal or such, just an ordinary bankruptcy, but he could demonstrate that it was impacting his current business.
He sued, and google had to remove the thing. Not when you search for his name and bankruptcy, not when you search for "what happened to that restaurant", and the newspaper itself also didn't have to do anything. As far as I know you can still find the article.
If you're a journalist writing the guy's biography, you'll find it, push come to shove in some offline archive. But random people won't see him nailed to a virtual pillory, that's what all this is about.
I don't think it's really an issue for AI, but it has to be engineered in. Ultimately it's about judging relevancy.
NixOS. Gentoo gets into trouble when you need multiple versions of the same library at the same time. Also while the infrastructure supports it it's annoying that gentoo doesn't provide pre-built binaries, like yes some people might want to have reason to build their own bash but I think I'll be fine with a standard build. NixOS? If you install something usually it's going to be pre-built. Change a couple of configuration flags? May or may not be pre-built. You want to apply a custom patch? It's going to seamlessly build from source.
Permanently Deleted
I don't know whether you've been alive during the "games make people violent" craze, where the establishment was trying to pin school shootings on a harmless hobby: The same establishment which caused the systemic issues driving youth into desperation was now blaming an escapist hobby for the issues they, themselves, caused. It's like framing jugglers for knife attacks. Do some jugglers juggle knives? Yes. Does that have anything to do with anything? No.
Is that in any way comparable in scale and intensity to what various groups had to go through historically, or are now, no of course not. But gamers, as a wider culture, know what it's like to be victim-blamed. The average person who picks up playing games as a hobby is not any more racist, misogynist, or drivelling rape apologia than the population average. They do it because they enjoy it, no further qualifications necessary.
But OTOH, nope. The copypasta is pumped-up to 11 when it comes to the capacity of gamers to be a community, to organise, take joint action, be a force, everything. I'd be easy to go "no true gamer" on some people but face it "gamers" as a group are still pre-ordering games and letting GPU scalpers have their way with us. Just as powerless as, say, football fans are powerless in he face of FIFA corruption. Concert goes in the face of ticket platform monopolies. We're a hopeless case in any other aspect but whinging: That, we are masters of.
I would guess it’s pretty hard to reconstruct data with a different LLM
I think the idea is to have compressor and decompressor use the exact same neural network. Looks like arithmetic coding with a learned function.
But yes model size is probably going to be an issue.
So... crystal ball, I don't have access to the paper either. Think arithmetic coders as neural nets are function approximators. You send an initial token and the NN will start to generate deterministically, once you detect a divergence from the lossless ideal you send another token to put it on track again. Make it a sliding window so things don't become too computationally expensive. You architect the model not to be smart but to need little guidance following "external reasoning" so to speak.
The actual disadvantage of this kind of thing will be the model size, yes you might be able to transmit a book in a kilobyte (100x or more compression) but both encoder and decoder will need access to gigabytes of neural weights, and that's just for text. It's also not going to be computationalliy cheap, though probably cheaper than PAQ.
Of course Germany won't extradite we don't extradite nationals to non-EU countries. It can even happen that we don't extradite Americans to the US because they can demonstrate that they're likely to face torture in the US, such as isolation cells.