It's the dot-com hype all over again.

Revolutionary, life-changing, grand promises. Insane investments in both capital and real value (land, hardware, etc.).

And in the end we'll end up with a few useful tools, a small number of successful companies, and a whole lot of dashed hopes.

Unless the slow destruction of the Earth's ecosystem worries you a bit. Because all of this stuff uses up resources. And while there is tons and tons of water on Earth, actual drinking water is already scarce in many parts of the world.

Not to mention that they own most of the land, which would free up that land for development such as building more AI data centers.

Win-win all around !

Yes, me personally. But not population globally. And any attempt to make laws regarding that (outside China) is considered a violation of the most basic human right.

It's not one of those "you can start with yourself" things. On the contrary, in most developed countries population is already shrinking. Which weirdly makes our politicians think they need to import people from countries that still breed like rabbits.

That's why I asked if it's a Windows thing.

The last time I actually switched off my iMac was over a year ago when I went on holiday for 2 weeks.

That's fair.

The 2.(stable/devel).patchlevel format worked extremely well and stopped version number explosions. The main drawback to it was that it was prior to git, and so the patchlevel could get very high. We also don't need stable/devel, any more, as we've now got one tree for stable and a different tree for devel.

Having said that, I did very much like the three digit split, even though (as Linus as repeatedly said) it was something of a fiction at times. We do sort-of have that, now, with the third digit being used to mark backported stability fix rounds. And, yes, I would agree that version numbering is a fiction of sorts anyway.

I really don't like the major number incrementing at the speed it does, though. Yeah, 3.5 years between a major number increment is sort-of ok. That's 42 months, and 42 is indeed the answer to life, the universe, and everything. And an OS kernel isn't. in all fairness, really susceptible to being divided up into the major.minor.patch format because none of these really mean anything in this sort of a context.

Dunno how you'd really go about improving the system.

Linux removes a feature from a kernel under one of two conditions only:

1. The feature isn't maintained any more AND is now so stale it cannot compile AND nobody is willing to take on the work to make it work
2. The feature refers to hardware that is so obsolete that the number of users is effectively zero insofar as anyone is capable of determining

As a result, you're generally safe with anything that is built into the official Linux kernel tree. The API provided to applications is incredibly stable and Linus reputedly has an army of dedicated berserker Vikings enforcing this.

However, binary-only drivers and non-standard components are another matter, as they're maintained out-of-tree and don't always comply with Linux kernel practices. This is the only area you have to be careful, as distros aren't always clear as to what is official and what is stuff they've grabbed off the net and linked in.

Indeed. One of the key areas in which AI research has lots of opportunity to advance is in efficiency. In theory, silicon-based intelligence should be both faster and more energy-efficient than neuronal intelligence because our neurons are actually quite slow and not terribly efficient. But the human brain's architecture obviously makes vastly more effective use of the processing power it has, so there's enormous opportunity for improvement in our current silicon analogs.

It will crack me up if after a few years and a few trillion dollars of investment into massive data centers for AI we make a big architectural efficiency breakthrough that makes the technology 3-4 orders of magnitude more efficient and renders all of those big data centers redundant. Perhaps we'll build out huge renewable energy production capacity to support AI and then suddenly find all of that capacity freed up for other purposes. Enormous green energy surpluses could power the carbon recapture needed to truly fix global warming as well as high-volume desalination and water pumping to fix water shortages... and lots more. That is, if AI doesn't kill us all.

Of course, we all know that half of global problems (climate change, pollution, too much energy usage, etc.) would disappear if half of the human population would vanish. But without Thanos, it's not like half of us would volunteer, right?

So, we don't have control over how many people there are. We DO have control over how much electricity we feed into AI systems.

Is physics just hubris, thinking you can explain things that are actually stranger than you are capable of imagining?

My ZX Spectrum booted in the time it took to flash the screen (less than 1 second I'd say).

If you had the Interface 2 it could boot directly into a game (via ROM chips that overrode the memory bus to present their data), so even faster.

Framework laptop.

Not long at all.

The restart/reboot is ridiculously fast.

Resume from suspend/hibernate is ridiculously fast.

The BIOS transferring to the bootloader? Seconds.

Honestly, it's like being in the year 2000 again. And my computer does what I say. Mostly because it's Linux.

Never cared for uptime on my desktop machines. My servers, on the other hand... let's just say it's how I found out that the kernel counter is 32bit. :-)

In sleep mode, it draws next to no power. In fact, I would argue that starting it up might eat up more power than leaving it in sleep mode. (I'm talking a few hours, over night or so, not for weeks or months, of course)