V0ldek

Correct answers are correct answers. The only thing LLMs typically are bad at, are things that are seldom discussed or have some ambiguity behind them.

Lol what, how many questions you ask in your life are entirely unambiguous and devoid of nuance? That sounds like a you issue.

but i still think that it’s a little suspect on the grounds that we have no idea how many times they had to restart training due to the model borking, other experiments and hidden cost

Oh ye, I totally agree on this one. This entire genAI enterprise insults me on a fundamental level as a CS researcher, there's zero transparency or reproducibility, no one reviews these claims, it's a complete shitshow from terrible, terrible benchmarks, through shoddy methodology, up to untestable and bonkers claims.

I have zero good faith for the press, though, they're experts in painting any and all tech claims in the best light possible like their lives fucking depend on it. We wouldn't be where we are right now if anyone at any "reputable" newspaper like WSJ asked one (1) question to Sam Altman like 3 years ago.

Okay I mean, I hate to somehow come to the defense of a slop company? But WSJ saying nonsense is really not their fault, like even that particular quote clearly says "DeepSeek said training one" cost $5.6M. That's just a true statement. No one in their right mind includes the capital expenditure in that, the same way when you say "it took us 100h to train a model" that doesn't include building a data center in those 100h.

Beside whether they actually lied or not, it's still immensely funny to me that they could've just told a blatant lie nobody factchecked and it shook the market to the fucking core wiping off like billions in valuation. Very real market based on very real fundamentals run by very serious adults.

It also takes literally 1.5s to search and find out what it was

I didn't read it because I don't think there's much emphasis on it in school outside of the anglosphere, but the 2005 movie was a classic, must've watched it a dozen times. Now that I recall who the director was, though, I kinda understand why you don't talk much about it anymore...

don’t even get me started on “whole language learning” and “new math”

I don't know what "whole language learning" is, and I'm way too young to have experience it, but wasn't the curriculum before "new math" like arithmetic and nothing else? In other words, not math at all?

I didn't read much into it but from what I did it seems like they started teaching children actual math like algebra and logic and parents got frustrated because they were too stupid to help with homework anymore. Brings into my mind the whole "math was cool before they involved letters" thing that makes me want to throw a book at someone.

dan brown’s shitty novels about the dude who deciphers symbols or whatever (it was the one with anti-matter)

Ah yes, litrtuere

"Aligning people is hard too" a thing that only a literal sociopath would think and only a special kind of sociopath would utter publicly

This is a really weird comment. Assembly is not faster than C, that's a nonsensical statement, C compiles down to assembly. LLVM's optimizations will most likely outperform or directly match whatever hand-crafted assembly you write. Why would BEQ 1000 be "considerably faster" than if (x == y) goto L_1000;? This collapses even further if you consider any application larger than a few hundred lines of code, any sensible compiler is going to beat you on optimizations if you try to write hand-crafted assembly. Try loading up assembly code and manually performing intraprocedural optimizations, lol, there's a reason every compiled language goes through an intermediate representation.

Saying that C# is slower than C is also nonsensical, especially now that C# has built-in PGO it's very likely it could outperform an application written in C. C#'s JIT compiler is not somehow slower because it's flexible in terms of hardware, if anything that's what makes it fast. For example you can write a vectorized loop that will be JIT-compiled to the ideal fastest instruction set available on the CPU running the program, whereas in C or assembly you'd have to manually write a version for each. There's no reason to think that manual implementation would be faster than what the JIT comes up with at runtime, though, especially with PGO.

It's kinda like you're saying that a V12 engine is faster than a Ferrari and that they are both faster than a spaceship because the spaceship doesn't have wheels.

I know you're trying to explain this to a non-technical person but what you said is so terribly misleading I cannot see educational value in it.

It's actually "Reasonings General", common misconception

AGI is coming, we're already at the "dumb guy who doesn't understand math but thinks he's smart" level

as it’s apparently no longer safe for trans people or C++ developers

Sorry but Rust knowledge is now a hard requirement for visas so you better hit the book