jjaksic

Have you tried Bazzite or Aurora? A well-made and full-featured atomic distro is basically as easy to use and as reliable as ChromeOS, but with most of the flexibility that Linux offers.

I had Ubuntu for 8 years and it was as you describe (things broke all the time and every time I had to spend hours searching forums for arcane command lines etc). With Bazzite everything just works and nothing ever breaks. KDE with Wayland looks and feels amazing. I love love love the experience!

Do readability and maintainability not matter when AI "reads" and maintains the code? I'm pretty sure they do.

When Nadella took over from Ballmer, he steered Microsoft in a better direction for a while. But by now he's become a lot worse. The biggest software company can no longer produce good software and its products are actively hostile to users. Nadella cares only about one thing, which is shoving AI everywhere and to everyone, at any cost. The irony is that he knows nothing about AI, how to build capable models or how to build useful AI products, nor does he have people who do. AI is his Metaverse: something he's singularity focused on, to the point of neglecting everything else, without any idea what to actually do with it.

"I'm going to go with Fedora on my tower and Bazzite (or SteamOS) on my handhelds."

Why not Bazzite on both? Bazzite is a fantastic desktop OS! Easier to use than naked Fedora and virtually unbreakable.

Seriously, one random website getting less traffic means "Google is dead"? I imagine if you hit your toe, you call it "end of the world"? This sort of posts should be illegal. Flagged.

FWIW, I once complained about this and they told me it's a "hard problem".

Linux has made an insane amount of progress in recent years. Atomic distros like Bazzite and Aurora are so polished, modern, easy to use, and virtually unbreakable. Even most Windows games work perfectly out of the box (often better than on Windows). Anyone who tried Linux in the past and wasn't happy, should take another look. These distros are so incredible it's hard to believe.

Meanwhile Windows has been getting worse and worse. Completely unreasonable and unnecessary hardware requirements, spyware, constantly running antivirus and other processes you don't want, forced updates and reboots, shoving AI down your throat. In other words, you pay money to have a worse experience and less control over your own PC.

I've been ideologically opposed to Windows for a while, but a few years ago Linux required many trade-offs and compromises, to the point I wouldn't have recommended to most people. But now things are completely different and I would happily recommend it to anyone except those who have a hard requirement for MS software (or Adobe).

Bazzite is an insanely awesome distro!

I used Ubuntu for 8 years constantly fixing issues, from the day I installed it (because it didn't support basic Ryzen), after every distro upgrade, and various other random points, e.g. when installing a package whose dependency overwrote something. Each issue took hours to fix, usually searching forums for arcane command lines and trying everything until something worked (possibly breaking other things in the process).

Last year I tried Bazzite for my kid who like games and realized that it's 100x better than Ubuntu, for both gaming and serious work. It's 100x more stable and virtually unbreakable, far more modern and up-to-date than Ubuntu, and I can still do just about everything I want (just have to do it differently because it's atomic). Since I switched to Bazzite I have had zero issues, because atomic distros are inherently so much more stable. Everyone uses the exact same image, and the state of the OS is always fresh and doesn't deteriorate over time the way mutable distros do. And best of all, if any issue does come up (which is extremely rare), the fix is always the same and it takes 1 minute: boot into the previous version.

I used to avoid using my PC due to constant issues with Ubuntu, now I often switch it on simply because Bazzite makes me so happy.

It frustrates me to no end that people to this day still recommend Ubuntu and its derivatives as "good" and "user friendly" when it literally breaks all the bloody time, and meanwhile there are awesome distros like Bazzite and Aurora that are rock-solid like MacOS and ChromeOS.

"I'm leaving you for a new context window."

Yes, exactly. AI has its uses and can sometimes be extremely useful. But at this point it's not nearly as ubiquitously useful as various companies would want us to believe, based on how much they're forcing it on us, pushing it in our faces, shoving it down our throats, etc. I don't want that. I'll use it if and when I want to, thank you very much. Microsoft is of course the worst offender.

"Hop on a call" is also tone-deaf.

You know you can sell and replace your phone if you don't like it. Recent Pixels have 7 years of support and they don't die. That's what I'd recommend you get instead. You can even trade in your iPhone for up to $700 when you buy a Pixel. You really don't need to force yourself to use a phone you don't like, leave alone for that long.

Because human work is designed for humans. If you want a drop-in replacement for human workers, humanoid robots are your best bet.

Wow, I don't understand what anything means in those memes. And I'm so glad I don't!

It seems to me that the people/committees who built C++ just spent decades inventing new and creative ways for developers to shoot themselves in the foot. Like, why does the language need to offer a hundred different ways to accomplish each trivial task (and 98 of them are bad)?

Being able to cut 90% of code sounds like someone was getting paid by LoC (which is also a practice from a time when C++ was considered a "modern" language).

Dude, that is so weird!

Linus is not your boss in the sense that he pays you and can tell you what you do day o day, but he is your "boss" in the sense that he's the one who ultimately approves your work (which includes both your code and your conduct).

It's a "two-way street" (you can walk away as much as he), but you need to understand that this is not an equal relationship. It might have been if Linux did not yet have a file system and you were the only person who could build one. If that was the case, Linus might have swallowed his pride for the good of the project. But as it stands, Linux has existed for some 35 years without Bcachefs and can continue to do so. So the key stakeholders have simply decided that with this amount of friction, despite the technical advantages, it's not worth the trouble (yes, it is sad).

Very realistically and bluntly, you have 3 options: a) Learn to work Linus and other kernel maintainers in ways that are comparable with their work and processes. Remember, you're on their territory, and when in Rome, do as the Romans do. b) Keep and develop Bcachefs out of kernel. That way you can stay on your own turf and work on your terms, but Bcachefs is going to be a much less attractive and viable option, leave alone become the default fs. c) Have someone else do the integration work and collaboration with the kernel team.

These are your options, but a combination is also an option. I would probably recommend starting with option b first and finish the bulk of the work out of tree. Then once it's all done and ready to ship, try to get it into the kernel as politely and timely as possible (you shouldn't need any late commits this time). Continue developing new and experimental functionality out of tree to keep the number of PRs (and thus possible causes of friction) low. I don't know if at any point you'd want to ask someone else to interface with the kernel team. I think it's far preferable if you can learn to do it yourself. Knowing how to work with people (including difficult people) is an incredibly important and useful skill in almost anyone's career. Maybe find a communication/diplomacy mentor? I've never heard anyone complain about your engineering skills, but this is really holding you back.

Again and as always, thank you for your hard work. I wish you all the best and hope that some day we can all use Bcachefs by default.

I was there. They had no idea. The purpose and expectation of the transformer architecture was to enable better translation at scale and without recurrence. This in itself would have been a big deal (and they probably expected some citations), but the actual impact of the architecture was a few orders of magnitude greater.

Illusion doesn't imply it's unnecessary. Humans (and animals) had a much higher probability of survival as individuals and as species if their experiences felt more "real and personal".

Introspection is just a debugger (and not a very good one).

As a physics layman I sometimes watched Sabine's show and found it interesting. The one where she defended Weinstein was the one where she lost all credibility to me and I stopped watching her.

Her (expletive-laden) message was essentially: "Weinstein is my friend. Yes, his theory is bullshit, but so is all of theoretical physics." Seriously, aren't you one of them? You would rather throw your entire academic field under the bus to defend your friend? (And mind you, what a great way to defend your friend, calling his theory bullshit.)

This blog post is incredibly illuminating and explains a lot. It's a prime example of "Don't expect someone to understand something when their YouTube paychecks depend on them not understanding it", a.k.a. audience capture.

It's also an important reminder of the precarious situation laypeople are in - being unable to tell what's true and what's bs, and often relying on social cues like how confident someone sounds. We are all laypeople in most fields and are subject to easy manipulation by various confident-sounding grifters and LLMs.

You should tell them.

Firefox has been my browser since its first beta 20-something years ago, and I don't see anything replacing it. It has by far the best addons and tab management. Chrome, after all these years, doesn't even offer vertical tabs, and its extensions API keeps getting more restrictive.

That seems fair and accurate.

I guess disagreement about this question often stems from what we mean by "human", even more than what we mean by "intelligence".

There are at least 3 distinct categories of human intelligence/capability in any given domain:

1) average human (non-expert) - LLMs are already better (mainly because the average human doesn't know anything, but LLMs at least have some basic knowledge),

2) domain expert humans - LLMs are far behind, but can sometimes supplement human experts with additional breadth,

3) collective intelligence of all humans combined - LLMs are like retarded cavemen in comparison.

So when answering if AI has human-level intelligence, it really makes sense to ask what "human-level" means.

There's a lot of moving the goal posts here. What would it mean for an AI to be considered "intelligent"? At various points it meant different things (chess, jeopardy, go, human language, math, etc). The most well-established benchmark was the Turing test, which LLMs now pass spectacularly -- in fact the only way to reliably tell that you're talking to an AI is usually because it responds too fast and too perfectly.

Sure, AI often makes mistakes, but so do humans. Ask a random person a random question and I guarantee that at least 90% of the times an LLM will produce a better answer. So if knowledge or reasoning mistakes are a benchmark, then humans are not intelligent either.

What is the difference then? I feel that whoever claims that AI is not intelligent simply feels uncomfortable thinking that they are, and is going out of their way to rationalize that, by using various double standards. Like, AI making occasional mistakes counts as evidence against it, but the fact that the vast majority of people are incapable of answering some of the most basic questions somehow doesn't count against human intelligence.

And then there are various contrived "definitions of intelligence", such as "living in the physical world". These have nothing to do with cognitive intelligence and are just made up to make humans fit the definition and AI to not fit. Doing "truly novel things"? How many people have ever come up with anything truly novel? What about the other 99.99%, are they intelligent?

I think there's one thing that's tripping people up. Humans do have a very high COLLECTIVE intelligence. Collectively (and given enough time) we're capable of accomplishing great things, much more than current AI. But each individual human on their own is no more intelligent than a caveman (we have essentially the same brain).

And what truly novel things are humans capable of? At least 99% of the stuff we do is just what we were taught by parents, schools, books, friends, influencers, etc.

Remember, humans needed some 100, 000 years to figure out that you can hit an animal with a rock, and that's using more or less the same brain capacity we have today. If we were born in stone age, we'd all be nothing but cavemen.

What genuinely novel thing have you figured out?

But "shortening other summaries from its training set" is not all an LLM is capable off. It can easily shorten/summarize a text it had never seen before, in a way that makes sense. Sure, it won't always summarize it the same way a human would, but if you do a double blind test where you ask people whether a summary was written by AI, a vast majority wouldn't be able to tell the difference (again this is with a completely novel text).

If you lock a human in a prison cell without a clock, so they can't tell the time and can't "go to a store", do they stop being intelligent?

My wife and I have 2 teslas, a HW3 and a HW4. Even late last year FSD 12.5 was nowhere near close to being able to drive safely. Any non-straightforward situation (like merging during rush hour) would throw it off, so critical interventions were required at least daily.

Starting with FSD 13 on HW4, which came out last December, it's improved dramatically, and since then in my case it hasn't needed a single critical intervention. I think 12.6 on HW3 is also quite good.

The caveat is that we live in the Bay Area, which has an abundance of Tesla training data. Elsewhere I've heard the experience isn't as good. And of course, even in the Bay Area the reliability needs to get a few orders of magnitude higher to be suitable for fully unsupervised self-driving.

This is somewhat tangential: people are debating a lot whether porn has negative effects, but has anyone thought about or studied positive effects? I personally think it's more likely for porn to have positive effects and that it's preferable to sexual deprivation. Horny and sexually deprived people tend to sometimes do awful things like sexual harassment or even rape, but if they have access to porn, they're much less likely to do those things.