Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

DeepSeek v4.1 Flash

twitter.com

Post reply on HN
Print view View original

481–490 of 514 posts

Re: DeepSeek v4.1 Flash

#481

Post by idiotsecant »

mrob wrote: post #365
idiotsecant wrote: post #357

Earlier quoted context omitted.

If I go to sleep, wake up, and then go to sleep again am I a different conscious entity each time?

Possibly. If I had some side-effect-free means to permanently prevent all sleep I'd take it without hesitation. But that's not relevant to the discussion, because it's not anything similar to what an LLM does. Your brain changes state even while sleeping.

In what way does your brain change state while sleeping that a model does not change state via constant fine-tune updates?

Re: DeepSeek v4.1 Flash

#482

Post by idiotsecant »

hgoel wrote: post #407
idiotsecant wrote: post #340

Earlier quoted context omitted.

If I operate a nuclear reactor or a hydroelectric dam there are regulators that tell me what i'm allowed to do, so as to keep my profit motive from overwhelming the public interest. If we want AI to actually have some safety rails, this is what we would do.

If we were to take the nuclear analogy, what's happening in AI right now is where the people selling nuclear power make a ton of noise about how they need the power to regulate their competitors because nuclear bombs might set the atmosphere on fire, and their proof for this is in a report about how they didn't wear TLDs (and suffered related increased cancer risks) despite it being common practice to wear them in al…

Regulations obviously can't be written by those regulated or they're just moats by another name.

Re: DeepSeek v4.1 Flash

#483

Post by sixothree »

azath92 wrote: post #187
tarruda wrote: post #133

Earlier quoted context omitted.

200B of those 552B is PLE, which works more like a database that is read for each token, thus can be offloaded to a fast SSD.

Id love an ELI5 for PLE. Im trying to work it into my back of the napikin math for compute vs memory bandwidth limitations on tok/s in PP vs TG work. My attempt at a simplification of this article on it https://sebastianraschka.com/llm-architecture-gallery/per-la... into a couple of sentences is that they are linear embeddings of the input token space projected per layer, which are then gated by the transformer outpu…

While he avoids using the actual PLE acronym, he does actually describe the concept quite well. I think you may enjoy this video. Specifically around 5 minutes into the video is the part you're looking for.

https://www.youtube.com/watch?v=1--PzaHafAU

Re: DeepSeek v4.1 Flash

#485

Post by lowbloodsugar »

walrus01 wrote: post #54

Looking at the huggingface page, the unsloth people haven't finished quantizing it yet, but I'm sure they're active on it right now. It'll be interesting to see how the capabilities and benchmark tests compare on system where it can fit in under 512GB of RAM with full context. In terms of coding and command line capabilities I'm also very interested to see a head-to-head of it vs. qwen 3.8-flash-next Q8 which is some…

3.8-flash-next fits on a single 6000 at Q4 if you offload the PLE. Crazy fast and still effective.

Re: DeepSeek v4.1 Flash

#486

Post by adwn »

IhateAI_6 wrote: post #202
nazgob wrote: post #109

Earlier quoted context omitted.

You compare Norwegian treatment of immigrants to Chinese Uyghurs?

You don't know anything about how the Chinese treat Uyghurs other than what you're told through western propaganda that you then repeat like you're pavlovs dog whenever the word gets brought up. Yes, its quite similar. Norway forces people to attend cultural schooling, where theyre provided housing in the interim. Its quite similar. However, the migrants in Norway are there because the Norwegian state through its par…

> Norwegian state through its participation in NATO murdered millions of people in the middle east

I'll file this under "blatant lie" unless you can provide at least three reliable sources for your claim.

Re: DeepSeek v4.1 Flash

#487

Post by Certhas »

fwn wrote: post #472
Certhas wrote: post #403

Earlier quoted context omitted.

Nonsense. There was one proposal for relevant quantum effects in brain dynamics, and that turned out to be not relevant. Even if they were, you could substitute all quantum randomness with pseudo randomness and obtain an absolutely indistinguishable object. But even if this were a debate, its absolutely absurd to claim that the question of determinism in the brain has any bearing on our moral standing. If we discover…

You seem to have conceded the "is it deterministic" argument only to sidestep by declaring determinism irrelevant. Your original claim was that LLMs are deterministic "in the same sense" as brains. We can write down an LLMs full register, and that register/book contains the whole output universe of the text generator. That book does not act, it is morally neutral. That the brain has such a register at all is just res…

I am a physicist. I worked (briefly) on foundations of quantum mechanics. I discussed compatibilism extensively with philosophers.

I have _never_ come across the position you seem to take here, that determinism has bearing on the question if we are conscious and sentient.

Re: DeepSeek v4.1 Flash

#488

Post by vuldin »

toasty228 wrote: post #474
Kuyawa wrote: post #398

DeepSeek Harness, install it, thank me later. You won't believe the productivity gains for just pennies https://deepseek.com/harness/en/

It's web only? No cli?

DSH also includes a TUI (but I have not used it yet). It's in heavy development also, so there are new versions almost daily.

Re: DeepSeek v4.1 Flash

#489

Post by browningstreet »

eile23 wrote: post #316

Has anyone tried this for coding yet? I'm curious how it compares to Claude or GPT models on larger codebases, especially for debugging and making changes across multiple files.

I cross code and review between Deepseek v4 Flash and Claude Opus/Fable. I will do a full code review of my project with DS 4.1 F later today, but Claude makes a lot of mistakes that DS finds. Claude is slightly more ambitious about what it's reaching for, but DS is far more proficient and efficient with a slightly lower ceiling. That may not be true after the upgrade, but if I had to live with one, as I'm paying out of my own pocket, I'd def stick with Deepseek.

Re: DeepSeek v4.1 Flash

#490

Post by reverius42 »

muddi900 wrote: post #447
Certhas wrote: post #121

Earlier quoted context omitted.

What's your definition of sentient? Or, maybe more precisely, consciousness? I think it's reasonable to at least start thinking about these questions. It has long been established that LLMs have good theory of mind [1]. And there is a bunch of empirical research about all sorts of capabilities that we typically associate with consciousness [2], like identity [3] and metacognition [4]. The METR report shows agents sac…

The same mechanisms exist in every machine learning algorithm. Is the spell check in Word conscious? Is the generative fill in photoshop conscious? A plane flies. It is much better at flight than bird(in terms of transportation). Is a plane also a bird?

I think you've got it backwards. The people arguing Claude can't be sentient because it's not human are arguing that a plane can't fly because it's not a bird.
Post reply on HN