𝙲𝚑𝚊𝚒𝚛𝚖𝚊𝚗 𝙼𝚎𝚘𝚠

Indian

I think the point is that even if LLMs suck at task A, they might be really good at task B. Just because code written by LLMs is often riddled with security flaws, doesn’t mean LLMs also suck at identifying those flaws.

AFAIK Taler is for payments, the Digital Euro is for storing value.

I remember a project where someone booted Linux off of Google Drive. Cursed on many levels.

True, there were several programming mistakes that caused undefined behaviour. Most of these the compiler warns about though, so they could have easily been fixed.

The issues were “masked” so to speak by the debug build (even if not fully gone, the game could still crash). But decompiling the game let modders fix those issues fairly easily, after which it could be recompiled with the proper optimizations.

Yup, they shipped a debug build. Here’s a video that shows the build side-by-side with one that was compiled with compiler optimizations: https://youtu.be/9_gdOKSTaxM

It was quite laggy in certain areas, particularly the submarine sank the framerate quite considerably.

Meh, those are just the programmers that are remembered.

They did lots of dumb shit too. Mario 64 was a super-innovative game at the time with its free 3D platforming. There’s also tons of weird code in there, and the developers also fucked up by shipping a debug build of the game, costing a not insignificant amount of performance.

From the FAQ, they want to eventually move to https://code.europa.eu/

Misspelled “Invidious” there :)

Looks neat though!

Deepseek seems to have done a clever thing w.r.t. training data, by having the model train on data that was emitted by other LLMs (as far as I’ve heard). That means there is sort of “quality-pass”, filtering out a lot of the definitely bogus data. That probably leads to a smaller model, and thus less training hours.

Google engineers put out a paper on this technique recently as well.

It’s missing the rest:

Resynthesizer is a Gimp plug-in for texture synthesis. Given a sample of a texture, it can create more of that texture. This has a surprising number of uses:

• Creating more of a texture (including creation of tileable textures)
• Removing objects from images (great for touching up photos)
• Creating themed images (such as the Resynthesizer logo above)

Eh, those usually ain’t too bad. Runtime Blazor errors usually are a bit more annoying, sometimes requiring you to open up the intermediate compiled cs files.

Unreal pushes a lot of “hip tech” that supposedly improves performance, but often it turns out that many example cases are just really poorly optimised. With more traditional optimization techniques more can be achieved.

Unreal can perform really, really well, it’s just that it won’t by default. And many devs are too lazy to properly profile their games to figure out how to improve it.

This is a massive cry from “behaves like humans”. This is “roleplays behaving like what humans wrote about what they think a rogue AI would behave like”, which is also not what you want for a product.

I don’t think “AI tries to deceive user that it is supposed to be helping and listening to” is anywhere close to “success”. That sounds like “total failure” to me.

The tests showed that ChatGPT o1 and GPT-4o will both try to deceive humans, indicating that AI scheming is a problem with all models. o1’s attempts at deception also outperformed Meta, Anthropic, and Google AI models.

Weird way of saying “our AI model is buggier than our competitor’s”.

And the cybertruck’s structure stopped anyone outside it from getting hurt…

The reason nobody outside was hurt was because there wasn’t really anyone around to get hurt. The video shows a pretty sizeable explosion that would’ve likely killed someone standing close by.

I don’t think the Cybertruck did any better than other cars in that respect. Not worse either btw.

“Look, Python is way easier to use than other languages! Look how complex this easy task is in Python versus other languages like assembly and brainfuck!”

I’m not saying “do stuff in C it’s easier than Python”, but if I took e.g. C# then it’s also just two lines. That supports everything and is also faster than the Python implementation.

I mean, is it? I personally haven’t found Python using much less boilerplate. It’s possible, but you end up with something inflexible that’s hard to maintain.

Different network layer. MAC addresses are layer 2 iirc, whereas IPv6 is layer 3.