brucethemoose @ brucethemoose @lemmy.world

Posts

22

Comments

2,106

Joined

1 yr. ago

Late to the post, but look into SGLang, OP!

In a nutshell, it’s a framework for letting LLMs “fill in blanks” instead of generating entire replies, so you could script in rules as part of the responses as structure for it to grab onto. It’s all locally runnable (with the right hardware, unfortunately).

Also, there are some newer, less sycophantic DM specific models. I can look around if you want.

Is it a virus that affects the brain?

Yes! It’s called engagement optimization. And the worlds collective ignorance “don’t feed the trolls.”

It feels unsustainable, right? Like the value of of this tsunami of advertising has to be inflated, especially with bots/agents taking over traffic. People’s tolerance for junk isn’t infinite. At some point the illusion has to crack, and the advertising bubble will pop and burn the internet/app ecosystems down, hopefully…

I don’t see how any prediction can possibly hold any water since tariffs are totally unpredictable and in flux, and the commerce volume effects they have are huge (and dependent on those rates). Deficit reduction itself is good, but it’s also a regressive tax that directly hits the consumer base.

But I will say the US was basically founded to avoid tariffs, and got real good at smuggling, too.

Heh. systemd-boot's Windows entry broke for me over a year ago, and I haven't been able to fix it. I have to boot Windows through UEFI.

Depends on the application.

In some cases, it would be fantastic. But it’s clearly not a one size fits all, yeah.

Eh, I still bet it was really people browsing /new who downvoted it.

Honestly I get it, with how enshittified corporate portals/use is already, but still.

It's not bots, it just how local ML posts are on the internet.

I got banned from a Reddit fandom sub for the mere suggestion that a certain fan 'remaster' be updated with newer diffusion/GAN models. Apparently they weren't aware the original was made with Waifu2x... But unfortunately, anything tangential to tech bro AI is radioactive.

TBH 2-3 would be good, since each browser takes a monumental amount of effort/money to optimize and maintain.

Like, my best case somewhat plausible scenario would be Apple (and maybe some other vested interests?) merging Firefox and Safari into one open source effort that can keep up with Google (with Safari being a “branded” Firefox). There just isn’t enough money for a couple of open efforts to keep up with Chromium.

+1, though all this is a very unpopular opinion on most of the internet.

Local models are not capable of coding yet, despite what benchmarks say. Even if they get what you’re trying to do they spew out so many syntax errors and tool calling problems that it’s a complete waste of time.

I disagree with this. Qwen Coder 32B and on have been fantastic for niches with the right settings.

If you apply a grammar template and/or start/fill in their response, drop the temperature a ton, and keep the actual outputs short, it's like night and day vs 'regular' chatbot usage.

TBH one of the biggest problems with LLM is that they're treated as chatbot genies with all sorts of performance-degrading workarounds, not tools to fill in little bits of text (which is what language models were originally concieved for).

Seems like your really pondering “HTML should be conspicuously slow for such a widely-used standard,” right?

The answer is that modern browsers are complex and highly optimized rendering engines.

Read back through this blog: https://mozillagfx.wordpress.com/

But in a nutshell, there’s a lot of talk about how modern browser are analogous to tuned game engines, heavily relying on the GPU and all sorts of hacks to render HTML efficiently. “Compiler” doesn’t even begin to do them justice, and modern GPUs are a core part of getting a good browsing experience.

V8 is another good example, taking what was a notoriously slow language (JavaScript) and hacking out a fast JIT engine for it.

The “standard” is ostensibly the HTML spec and such, but in reality whatever Chrome (and Firefox/Safari) renders is the standard. Devs build around their strengths and quirks, basically, instead of the other way around.

Indeed. The blockchain provides no media hosting, no enforcement, I guess. It can mark something as owned (and require their private key to decrypt or whatever), but ultimately that ownership is as beholden to reality (read: arbitrary purseholders) as any other system. It’s just a record.

Pytorch Nightly: https://pytorch.org/blog/compromised-nightly-dependency/

https://www.theregister.com/2023/01/04/pypi_pytorch_dependency_attack/

Funnily enough I can't even post what it does without the Lemmy comment filter zapping me, but it tried to scrape accounts and passwords.

The malicious binary would upload files ranging in size up to 99,999 bytes and send the contents to a specified domain.

Was pretty scary from my perspective. I missed it by a week. PyPi is a mess, and it makes me wonder how much isn't caught.

There already are. I barely missed a linux virus from a hijacked python package what... two years ago?

Linux desktops are quite non-homogenous though, so their vectors/nature is kinda different.

I guess the problem NFTs try to solve is authority holding the initial verification tied to the video. If it’s on a blockchain, theoretically no one owns it and the date/metadata is etched in stone, whereas otherwise some entity has to publish the initial hash.

In other words, one can hash a video, yeah, but how do you know when that hashed video was taken? From where? There has to be some kind of hard-to-dispute initial record (and even then that only works in contexts where the videos earliest date is the proof, so to speak, like recording and event as it happens).

Funny thing is correct json is easy to "force" with grammar-based sampling (aka it literally can't output invalid json) + completion prompting (aka start with the correct answer and let it fill in whats left, a feature now depreciated by OpenAI), but LLM UIs/corporate APIs are kinda shit, so no one does that...

A conspiratorial part of me thinks that's on purpose. It encourages burning (read: buying) more tokens to get the right answer, encourages using big models (where smaller, dumber, (gasp) prompt-cached open weights ones could get the job done), and keeps the users dumb. And it fits the Altman narrative of "we're almost at AGI, I just need another trillion to scale up with no other improvements!"

Fissile material (enriched uranium and plutonium) has a long half-life.

AFAIK tritium is the biggest issue, with a half life of 12.3 years. From what I've read, nukes have little feeder tubes to replenish that.