I tried to use an LLM to write a script for me. It confidently told me I could split a string in OpenSCAD with the [1:] operator. It works in Python, but isn't an OpenSCAD feature.
Fortunately, programming has a good way of letting you know when the LLM is completely wrong.
The worrying part is that the LLMs can sometimes produce code that runs, but has massive security issues that you don't notice if you just run the code and don't closely analyse it.
It was the best option at the time, and I didn't feel like setting up Windows more than once as I gradually added new software and removed other software.
That particular VM doesn't get booted very often anymore.
I use it in a VM because it's better than Windows 10, some programs refuse to run on Linux, and I don't need to provide the VM with an internet connection.
My actual server is an old laptop and it does not run Windows.
Even my friend who insists that Windows is superior still uses Linux for his gaming server because it's easier to set up and manage.
If they get one angry message they'll ignore it. If they get 10,000 angry messages they might start worrying about what the people who haven't sent messages will do if they proceed.
Public outcry doesn't always work, but it has worked enough times to try again. It takes minimal effort for a significant potential gain.
Something similar to the way files (and bank transaction records) are modified on a drive. A first message indicating that a specific transaction is starting, a second message with the transaction, and a third message indicating the transaction is complete.
If any message is not acknowledged, it can be resent (with an ID so it's recognized as a duplicate), and if all else fails, flag a human.
The system doesn't need to be 100% accurate, just good enough that only a few humans are needed to fix the problems.
The Byzantine Generals problem is only a problem if you need 100% certainty from both sides. If you're willing to put up with 99.99% certainty, then network engineers have solved this problem many years ago.
If you have multiple banks, you can have multiple copies of the same database. If any one bank modifies their copy, it is obvious who modified it and what was changed. No blockchain needed.
The text requests aren't the (electricity) problem we've been complaining about. The never ending training is the big power consumer.
It's somewhat telling that Google went to a lot of effort to estimate the energy usage of text queries, but omitted any estimate of training energy usage.
I still won't touch a Tesla until Musk is removed, and they start supporting right to repair. Even then, their reputation for quality and customer support is hard to get past.
Is "main" a valid Java identifier?
package main; public class main { static main main; String Main; main(String main) { Main = main; } main(main main) { System.out.println(main.Main); } main main(main main) { return new main(main); } public static void main(main...Main) { main: for (main main : Main) { main = (main instanceof Main) ? new main(main): main.main(main); break main; } } public static void main(String[] args) { main = new main("main"); main.main(main, main); main = main.new Main(main) { main main(main main) { return ((Main)main).main(); } }; main.main(main); main.main(main,main); } abstract class Main extends main { Main(main main) { super("main"); } main main() { main.Main = "Main"; return main; } } }