kcb

That's why we all still use our e machines its never obsolete PCs. Works just the same it did 20 years ago, though probably not because I've never heard of hardware that's guaranteed not to fail.

My point is LLMs aren't more usable if the hardware is in your room versus a few states away. Personal computers still to this day aren't great when the hardware is fully remote.

The difference is computers post text terminal are latency and throughput dependent to the user. LLMs are not particularly.

That's 4 32GB GPUs with 600GB/s bw each. This model is not running on that scale GPUs. I think something like 96GB RTX PRO 6000 Blackwells would be the minimum to run a model of this size with performance in the range of subscription models.

Exactly the point though. In the 640KB days there was no subscription to ever increasing compute resources as an alternative.

Subscription services get all the same benefits from computer hardware getting better. But actually due to scale, batching, resource utilization, they'll always be able to take more advantage of that.

The 4-bit quants are 350GB, what hardware are you talking about?

What benefit is there to dropping $50k on GPUs to run this personally besides being a cool enthusiast project?

Nemotron 3 Super was released recently. That's a direct competitor to gpt-oss-120b. https://developer.nvidia.com/blog/introducing-nemotron-3-sup...

Victory is my device and its OS working the same way it always worked and the way it worked when I bought it.

Because the initial announcement included none of that... it wasn't addressed at all until the harsh sentiment.

Inference isn't really that expensive, its the training of new foundational models that is. With whatever highly optimized setup the big providers are using, they should be able to pack quite a lot of concurrent users onto a deployment of a model. Just think too, it's very possible their use case would be served just fine by a 100B model deployed to a $4,000 DGX Spark.

It's glass...

Another factor, it's not just GPUs it's the full hardware stack. https://static.tweaktown.com/news/1/1/110521_2_nvidia-update...

CUDA has had managed memory that pages between VRAM and system RAM for a decade. Problem is doing so is unusably slow for AI purposes. Seems like an unnecessary layer here.

But it has chemicals in it...

The 100 class Nvidia chips are targeted at training. With Nvidia buying Groq it will further move in that direction.

There's already videos of US/Israeli jets over tehran dropping guided bombs.

Mining is all compute and no IO. Training particularly is heavy compute and insane IO.

it's just HN getting baited for the 100th time by an electrek article.

Discord has parental controls. There's a myriad of services out there that restrict and monitor phone usage for kids. Use them and lock their phone and discord accounts down to nothing.

Restricting adults because parents decide to give little Timmy unrestricted access to technology is stupid.

> The biggest selling point /was/ that Musk was being managed there, he wasn't tinkering with SpaceX like Twitter or Tesla, and his foolhardy direction was kept out of the company

The biggest selling point to who? Definitely not wall street

And probably running on their macbooks...

How does that even work? What does it apply to? Say I own a 100% share in a business, each year does the government appraise it and pretty much require me to divest a portion of it to pay the tax?

Unrealized capital gains taxes are crazy all in an effort to own the rich or something. Meanwhile the people they're perceived as targeting have all the resources to avoid it.

My monthly plan payment would not go down if I brought my own phone.

Doesn't this depend on the application. For example electron applications dgaf about this system, render to a bitmap, and then look terrible as a result.

Crossover maintained compatibility with 32-bit windows applications even after Catalina.

> https://www.codeweavers.com/blog/jwhite/2019/12/10/celebrati...

Which is kind of funny because yet again windows was a better application in terms of longevity than MacOS native.

I was just thinking about this on my 60 mile FSD driver I just finished. Basically inevitable that I would shortly go HN or reddit and read how FSD doesn't work.

FSD is here, it wasn't 3 or 4 years ago when I first bought a Tesla, but today it's incredible.

The Raspberry Pi GPU has one of the better open source GPU drivers as far as SBCs go. It's limited in performance but its definitely being used for rendering.

The Linux Subsystem for Windows