Hacker Newsnew | past | comments | ask | show | jobs | submit | robotpepi's commentslogin

for sure they can do whatever they want, but that doesn't make it "pro consumer" as said above

[flagged]


you may want to re-evaluate your understanding of the word hackers. It has had a few definitions during my lifetime and none of the fit with how you have used it here.

My understanding of hackers is entirely based on people on this message board.

And of course hackers and other similarly minded people would be outraged if something they bought on Amazon could be found for cheaper elsewhere, and then the company would have to process a thousand percent increase in returns as well as losing the sales.


> Probably $500K worth.

What life standards do you have!?


they explain it in the article: this is the first iteration, so they wanted to start with something simple, ie, this is a tech demo.

Ok then I look forward to seeing DeepSeek running instantly at the end of April.

Why so negative lol. The speed and very reduced power use of this thing are nothing to be sneezed at. I mean, hardware accelerated LLMs are a huge step forward. But yeah, this is a proof of concept, basically. I wouldn't be surprised if the size factor and the power use go down even more, and that we'll start seeing stuff like this in all kinds of hardware. It's an enabler.

You don't know. You just have marketing materials, not independent analysis. Maybe it actually takes 2 years to design and manufacture the hardware, so anything that comes out will be badly out of date. Wouldn't be the first time someone lied. A good demo backed by millions of dollars should not allow such doubts.

Did you not see the chatbot they posted online (https://chatjimmy.ai/)? That thing is near instantaneous, it's all the proof you need that this is real.

And if the hardware is real and functional, as you can independently verify by chatting with that thing, how much more effort would it be to etch more recent models?

The real question is of course: what about LARGER models? I'm assuming you can apply some of the existing LLM inference parallelization techniques and split the workload over multiple cards. Some of the 32B models are plenty powerful.

It's a proof of concept, and a convincing one.


it's incredible how many people are commenting here without having read the article. they completely lost the point.

Are they becoming better at the same rate as before though?

In my unscientific experience, yes, but being better at a certain rate is hard to really quantify, unless you just pull some random benchmark numbers.

Per release, I’d say no.

Per period of time, I’d say yes.



yes, pretty much

this reminds me of that joke of someone saying "it's crazy that we have ten different standards for doing this", and then there're 11 standards

Xkcd 927

what kind of question is that

I cringe every time I came across these posts using words such as "humans" or "machines".

How would you call something like Claude or ChatGPT then, or even some image classifier from 20 years ago?

Just answering because I first wanted to write "software" or whatever.

I used to find gamers calling their PC "machine" hilarious.

However, it is a machine.

And for AI chatbots, I used the word for lack of a better term.

"Software" or "program" seems to also omit the most important part, the constantly evolving and intransparent data that comprises the machine...

The alogorithm is not the most important thing AFAIK, neither is one specific part of training or a huge chunk of static embedded data.

So "machine" seems like a good term to describe a complex industrial process usable as a product.

In a broad sense, I'd call companies "machines" as well.

So if the cringe makes you feel bad, use any word you like instead :D


I mean, you're right but at the same time you're talking about something completely different. Software with security vulnerabilities is not a useful product. You don't address the raised issues.

I'm completely ignorant about this, but wouldn't it be possible to compile separately your project to improve compilation times? for instance, if you're using OP's vector library, which is self contained, you could compile that first and just once?

Let's say you need to use a function like:

    int add(int a, int b){
        // Long logic and then this
        return a+b;
    }
Let's say this is your main.c.

    #include "add.h"

    int main(void) {
      return add(5,6);
    }

The preprocessor just copies the contents of add.h into your main.c whenever you're trying to compile main.c. (let's ignore the concept of precompiled headers for now).

What you can instead do is just put the add function declaration in add.h that just tells the compiler that add function takes two integers and returns an integer.

   int add(int a, int b);
You can then put the add function definition in add.c , compile that to an add.o and link it to your main.o at link time to get your final binary - without having to recompile add.o every time you change your main.c.

Precompiled headers: https://maskray.me/blog/2023-07-16-precompiled-headers


Unless you have link time optimization you would lose out on optimization and performance.

The whole thing is essentially a workaround for lack of sufficiently good/easy ways to package code in the ways people want to use it.


Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: