If you've got a toy project that you want "AI" to give you a hand with, do it now.
Pretty soon all these companies are going to have to pay for all that investment in compute resources they've been busily soaking up over the last few years, and then they're going to have to pay back their investors, and then they're going to have to try and make a profit
This is the golden time for cheap commercial AI. Already the noose is starting to tighten, and it will never again be as cheap as it is now.
Exactly. Functional public health systems will assess patient outcomes and the expenditure in money and resources to determine what treatments get approved.
The odds are pretty good that - if this works out - this will be on the list of approved treatments straight away. Surgery is an expensive and high-load pathway for public health systems. A non-surgical treatment that gives good outcomes is such a win-win for both patients and public health systems that it almost doesn't matter how much it costs.
To be honest I'd rather get OOM errors in my consumer OS than endless disk thrashing. At least mechanical drives were more resilient and you could hear the thrashing, as opposed to silent massive wear on your SSDs.
Yeah, it's on the to-do list. I still don't like it endlessly paging though. I also hate Microsoft's push to get everyone on win11 even if their PC's will run like asthmatic slugs.
No, the worst app is Windows 11. My parents have some all-in-one HP PC. It has some Intel laptop processor from about 8 years ago, 16GB of ram, (upgraded from 8), and a wheezy 256GB spinning drive, running Win10 adequately.
After being bombarded with "upgrade to Windows 11! It's easy and fun!!" notifications, they did so, and of course their PC is woefully underpowered for the job.
I log in remotely and check what's running and the OS is paging to the swap file constantly. [Edit: and I mean, to the point where opening an application takes 60 seconds of disk thrashing for a window to appear]
I had to get a de-bloat script and turn off about 50 Microsoft "essential services". The biggest hog out of them all was copilot, which was using about 4GB while sitting there idle.
I have no doubt that I'm going to have to run that script every month as everything gets "repaired", until I can get back to their place and put a SSD in and maybe install some flavour of Linux.
So does every city, suburb, town, factory, mine, processing plant, farm, etc.
What you're missing is that all these things provide economic benefits locally.
A datacentre uses prodigious amounts of local resources, but doesn't provide much in the way of local returns. It's very much a case of "privatise the profits, socialise the losses" for the locals.
Article talking about a little robot taking images of the lander, no images given.
So here's an image of the faceplanted lander that I found in a 5 second search, because Gizmodo couldn't be bothered to show it in the article, preferring to link to one of their other articles instead to get those juicy clicks.
I use them for a bit of coding leverage, and they require a fair bit of... not hand holding, but explicitness, maybe. Otherwise they wander down the path of statistical averageness, and the average code they saw during training was shit.
So they'll happily serve up a pile of inefficient dogshit, complete with working tests and docs and all of that. And then they'll happily refactor it at your suggestion to slowly turn it into something that's resource friendly and generally secure and generally expandable.
But there's no way for them to do that by themselves if you don't have the basic domain knowledge to guide them in the right direction. You'll just get average code, and once you've been in the game for a few decades, you realise that nice, efficient, quality code is the rarity, not the norm.
xml is better than markdown, you will not change my mind
Markdown is for humans who are writing their prose on the fly and don't want to be bothered with the cognitive load associated with the extra verbiage and rulesets of XML.
XML is for computers which are great with verbiage and rulesets.
It's perfectly fine to have a converter between the two to bridge the gap.
There was a good guide by Linuxbabe on building an email server from scratch with all the bits and pieces and antispam/email verification stuff you need to send mail to the big players, I used it a few years ago to do my server.
Here's the collection of various guides for various ways to do it:
Yeah you also need a vps. Home addresses are pretty much all marked as spam generators these days, and most ISPs proactively block all the common inbound ports for mail servers.
Remember kids, never update software when the change log just says, "bug fixes and performance improvements 😎👍🎉"
Although Meta has their own internal update system that invisibly runs outside the normal channels, because fuck those people who disable auto update on their devices, right?
Yeah, probably won't even dismantle it, just bulldoze this 10 billion dollars into a pit and bury the lot. Then go spend another 10 billion on a shiny new tent+compute
If you're not creating more than 800GB a day of new data you can just let it run with a faster drive as a buffer in front of it.
Or get 10 of them and run them in parallel. Maybe get 11 and throw in a bit of parity, just in case bitrot surfaces after the first 1000 years or something.
If you've got a toy project that you want "AI" to give you a hand with, do it now.
Pretty soon all these companies are going to have to pay for all that investment in compute resources they've been busily soaking up over the last few years, and then they're going to have to pay back their investors, and then they're going to have to try and make a profit
This is the golden time for cheap commercial AI. Already the noose is starting to tighten, and it will never again be as cheap as it is now.