Rose

Or: “The algorithm and data structure theory stuff is still pretty relevant. However, all of the examples are written in a language no one really uses any more. If they can get away with it.”

Methinks “rescheduled” is a more likely outcome.

Trump’s flunkies had 4 years to write Project 2025, so of course that shit has been implemented at frightening precision.

But they’re incredibly bad at doing anything on short notice. Expect Four Seasons level brainfarts if they need to improvise a solution yet again.

Alphabet starting with M? Hey, prescriptivist dingdongs, you can’t just say there is now apparently an alphabet that starts with M! Those things don’t exist in vacuum - if you make something like this up, you have to go all the way and define such an alphabet clearly!

Rules mean nothing unless they clarify established language use conventions! Which is why I think descriptivist approach is better in most cases.

I’m not opposed to AI research in general and LLMs and whatever in principle. This stuff has plenty of legitimate use-cases.

My criticism comes in three parts:

1. The society is not equipped to deal with this stuff. Generative AI was really nice when everyone could immediately tell what was generated and what was not. But when it got better, it turns out people’s critical thinking skills go right out of the window. We as a society started using generative AI for utter bullshit. It’s making normal life weirder in ways we could hardly imagine. It would do us all a great deal of good if we took a short break from this and asked what the hell are we even doing here and maybe if some new laws would do any good.

2. A lot of AI stuff purports to be openly accessible research software released as open source, and stuff is published in scientific journals. But they often have weird restrictions that fly in the face of open source definition (like how some AI models are “open source” but have a cap on users, which makes it non-open by definition). Most importantly, this research stuff is not easily replicable. It’s done by companies with ridiculous amount of hardware and they shift petabytes of data which they refuse to reveal because it’s a trade secret. If it’s not replicable, its scientific value is a little bit in question.

3. The AI business is rotten to the core. AI businesses like to pretend they’re altruistic innovators who take us to the Future. They’re a bunch of hypemen, slapping barely functioning components together to try to come up with Solutions to problems that aren’t even problems. Usually to replace human workers, in a way that everyone hates. Nothing must stand in their way - not copyright, no rules of user conduct, not social or environmental impact they’re creating. If you try to apply even a little bit of reasonable regulation to this - “hey, maybe you should stop downloading our entire site every 5 minutes, we only update it, like, monthly, and, by the way, we never gave you a permission to use this for AI training” - they immediately whinge about how you’re impeding the great march of human progress or someshit.

And I’m not worried about AI replacing software engineers. That is ultimately an ancient problem - software engineers come up with something that helps them, biz bros say “this is so easy to use that I can just make my programs myself, looks like I don’t need you any more, you’re fired, bye”, and a year later, the biz bros come back and say “this software that I built is a pile of hellish garbage, please come back and fix this, I’ll pay triple”. This is just Visual Basic for Applications all over again.

I’m from Finland. We like knives over here.

That’s entirely too many knives.

You kids have “serial” ATA? Ohhhh, watch this. Imma shoot these bits down this set of wires. Simultaneously.

“I don’t have the habit of looking at other people’s watches. Thanks for pointing it out, I guess, I might have otherwise missed that detail entirely.”

(Biz-bro mind cannot comprehend this)