6mo ago

Lol. Lmao even. "DeepSeek R1 reproduced for $30: Berkeley researchers replicate DeepSeek R1 for $30—casting doubt on H100 claims and controversy"

DeepSeek R1 reproduced for $30: Berkeley researchers replicate DeepSeek R1 for $30—casting doubt on H100 claims and controversy

Sam "wrong side of FOSS history" Altman must be pissing himself.

Direct Nitter Link:

https://nitter.lucabased.xyz/jiayi_pirate/status/1882839370505621655

77 comments

They finetuned 1.5-3b models. This is a non-story
- The headline is dumb, but the research isn't. According to the actual contents of the article, that $30 is still 27 times cheaper than what it costs OpenAI to make a similar sized model which also performs worse. That's still a big deal even if the people reporting on it made a stupid title for their article about it.
- I feel like the author here doesnt know what the definition of "breakthrough" is.
- Yup and it's not even testing general reasoning. They didn't have money for that.
  
  fuck almighty have these DeepSeek threads been attracting a lot of LLM “experts”
  
  Is General reasoning in the room with us now?
To reference a previous sidenote, DeepSeek gives corps and randos a means to shove an LLM into their shit for dirt-cheap, so I expect they're gonna blow up in popularity.
open source behaving like open source? couldn't be the evil scary chinese!
- open weights is not open source. If it were, then nobody would have to work on trying to reproduce it. They could just run the build script.
  
  unfortunately, nobody cares cos they're all thieves
  
  OSI is gonna mandate that we call it open source now, didn't ya hear?
Non-techie requesting a laymen explanation if anyone has time!
After reading a couple of”what makes nvidias h100 chips so special” articles I’m gathering that they were supposed to have a significant amount more computational capability than their competitors (which I’m taking to mean more computations per second). So the question with deepseek and similar is something like ‘how are they able to get the same results with less computations?’ and the answer is speculated to be more efficient code/instructions for the AI model so it can make the same conclusions with less computations overall, potentially reducing the need for special jacked up cpus to run it?
- Good question!
  The guesses and rumours that you have got as replies makes me lean towards "apparently no one knows".
  And because it's slop machines (also referred to as "AI", there is always a high probability of some sort of scam.
  
  pretty much my take as well. I haven’t seen any actual information from a primary source, just lots of hearsay and “what we think happened” analyst shit (e.g. that analyst group in the twitter screenshot has names but no citation/links)
  and doubly yep on the “everyone could just be lying” bit
- From a technical POV, from having read into it a little:
  Deepseek devs worked in a very low level language called Assembly. This language is unlike relatively newer languages like C in that it provides no guardrails at all and is basically CPU instructions in extreme shorthand. An "if" statement would be something like BEQ 1000, where it goes to a specific memory location(in this case address 1000 if two CPU registers are equal.)
  The advantage of using it is that it is considerably faster than C. However, it also means that the code is mostly locked to that specific hardware. If you add more memory or change CPUs you have to refactor. This is one of the reasons the language was largely replaced with C and other languages.
  Edit: to expound on this: "modern" languages are even slower, but more flexible in terms of hardware. This would be languages like Python, Java and C#
  
  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.
  
  for anyone reading this comment hoping for an actual eli5, the "technical POV" here is nonsense bullshit. you don't program GPUs with assembly.
  the rest of the comment is the poster filling in bad comparisons with worse details
  
  I’m sure that non techie person understood every word of this.
  
  Putting Python, the slowest popular language, alongside Java and C# really irks me bad.
  The real benefit of R1 is Mixture of Experts - the model is separated into smaller sections, that are trained and used independently, meaning you don't need the entire model to be active all the time, just parts of it.
  Meaning it uses less resources during training and general usage. For example instead of 670 billion parameters all the time, it can use 30 billion for specific question, and you can get away with using 2% of the hardware used by competition.
- The article sort of demonstrates it. Instead of needing inordinate amounts of data and memory to increase it's chance of one-shotting the countdown game. It only needs to know enough to prove itself wrong and roll the dice again.
- i read that that the chinese made alterations to the cards, as well-- they dismantled them to access the chips themselves and were able to do more precise micromanagement that cuda doesn't support, for instance.. basically they took the training wheels off and used a more fine-tuned and hands-on approach that gave them some serious advantages
  
  got a source for that?

77 comments