No. RSI doesn't mean a model modifies its own running code/weights. It trains another copy of them as a baseline for the next round of the experiment.

They are. It's just not happening on the versions we have access to. Look at Sakana. RSI just involves defining an experiment and parameters, and varying the values of the parameters, trying out a training run, and record the results. If it works well, use the results as part of the base model for the next iteration. All of that can be automated. I have no idea why people are saying they can't be. If people iterate on them and change one variable at a time, and use that as a means to improvement, an AI can be tasked to do the same. And they can also be trained to design the experiments themselves.

My point being that when you wrote "There's no GPS in my PC, idiots", the fact that there are no GPS does not mean your PC cannot locate you, and thus the popup could still be relevant.

Now is it annoying like hell? Heck yes. Not my point at all.

Regarding location data, even on a laptop or a desktop, an imprecise location can be deducted from wifi triangulation. You still need wifi though which is present on all laptops and many desktops sold these days. Or can be added with a USB wifi adapter.

I'm slightly more willing to give a pass to Xorg applications because it's not a monolithic environment supposedly under the control of a corporate entity that should be able to enforce its own standards.

of course you have to authenticate, but you run all of the software to administer it on a client, not sitting at the console. The console had only the most barebones capability, usually user management. The tools to administer ran on the client.

...it makes sense to have a headless server operating system when you're mostly running commodity spin-up/spin-down headless servers. Microsoft's server operating system was still largely based on the idea of running on a baremetal self-contained box, even though Microsoft servers had long, long since been used in the virtual machine space. If anything they're quite far behind the curve on this.

The Novell Netware model adapted to the VM era is what makes sense, where the tools don't require logging in to the server at all in order to administer the environment.

IBM is and has always been a services/consultant business, even when they made products.

I'm not so sure about the UI. The history of Microsoft and UI for the past 40 years is that they're happy to abandon their incumbent UI for different. We saw that with Windows 3.x to '95 and NT4, with Windows 98 and the integration of Spyglass Mosaic Internet Explorer, with the transition from Windows ME and Windows 2000 to Windows XP, the subsequent further transition from XP to Windows 7, and the rework from Windows 8.x to Windows 10. We even saw it with Windows 10 to Windows 11.

They change their UI because their customers don't see the OS being new/different unless they change their UI. If the UI looks the same then the average untrained end user doesn't know the difference and doesn't see a value in spending the money to upgrade.

Class rank exists and can be used to determine relative capability though.

I have no complaint with the idea that most students simply won't be able to achieve an A-grade if the material is both challenging and taught to proper standards, but I have a major problem with the notion that teachers are required to deny students that have mastered well above 90% of the material an A-grade because other students managed to yet outperform them. I hate the idea of grading on a curve. One should be judged against the mastery of the material, not comparatively against other students during that particular semester.

That said, I have also had college classes where I really should have failed the class but because of the curve, I got an A because I had the highest scores. While some of that reflects upon me, a good chunk of that reflects upon the instructor, the department and its head, and the curricula for that particular course. If students are to be held to high standards then instructors should likewise be held to high standards, and so should their institutions. If they cannot produce results then that should reflect both upon them and upon the revenue they receive in tuition.

And what about the computer that was used to write to the floppies?

Small teams running the show is the new normal. AI coding assistants mean that software engineers must also be product engineers, well enough to explain how to build features from soup to nuts, to LLMs. I know. this is what I do for a living now.

You can turn off auto-updates, but you'll be sorry in other ways.

When you copy files to your computer, that's after you have clicked 'Install' or whatever (or passed the -y flag on the CLI). That's your consent to install the app, and copy the files that come with it to your computer. There is no additional consent required. What the fuck are they even on about with this clickbait nonsense?