[HN Gopher] The Illustrated DeepSeek-R1
       ___________________________________________________________________
        
       The Illustrated DeepSeek-R1
        
       Author : amrrs
       Score  : 487 points
       Date   : 2025-01-27 20:51 UTC (1 days ago)
        
 (HTM) web link (newsletter.languagemodels.co)
 (TXT) w3m dump (newsletter.languagemodels.co)
        
       | caithrin wrote:
       | This is fantastic work, thank you!
        
       | jasonjmcghee wrote:
       | For the uninitiated, this is the same author as the many other
       | "The Illustrated..." blog posts.
       | 
       | A particularly popular one:
       | https://jalammar.github.io/illustrated-transformer/
       | 
       | Always very high quality.
        
         | punkspider wrote:
         | Thanks so much for mentioning this. His name carries a lot of
         | weight for me as well.
        
         | jamestimmins wrote:
         | Have you read his book Hands-On Large Language Models?
         | 
         | Looks interesting, but I'm skeptical that a book can feasibly
         | stay up to date with the speed of development.
        
           | jasonjmcghee wrote:
           | I have not, but Jay has created a ton of value and knowledge
           | for free and don't fault him for throwing an ad for his book
           | / trying to benefit a bit financially.
        
             | jamestimmins wrote:
             | Yeah no shade for someone selling their knowledge; I'm just
             | trying to suss out how useful the book is for learning
             | foundations.
        
               | whoisburbansky wrote:
               | Foundations don't change much with "the speed of
               | development"
        
               | jamestimmins wrote:
               | That's a good point
        
           | jampekka wrote:
           | > Looks interesting, but I'm skeptical that a book can
           | feasibly stay up to date with the speed of development.
           | 
           | The basic structure of the base models has not really changed
           | since the first GPT launched in 2018. You still have to
           | understand gradient descent, tokenization, embeddings, self-
           | attention, MLPs, supervised fine tuning, RLHF etc for the
           | foreseeable future.
           | 
           | Adding RL based CoT training would be a relatively
           | straightforward addendum to a new edition, and it's an
           | application of long established methods like PPO.
           | 
           | All "generations" of models are presented as revolutionary --
           | and results-wise they maybe are -- but technically they are
           | usually quite incremental "tweaks" to the previous
           | architecture.
           | 
           | Even more "radical" departures like state space models are
           | closely related to same basic techniques and architectures.
        
             | mistrial9 wrote:
             | > gradient descent
             | 
             | funny mentioning the math but not the Transformer
             | encoders..
        
               | jampekka wrote:
               | Transformer encoders are not really popular anymore, and
               | all the top LLMs are decoder-only architectures. But
               | encoder models like BERT are used for some tasks.
               | 
               | In any case, self-attention and MLP is the crux of
               | Transformer blocks, be they in the decoder or the
               | encoder.
        
               | mistrial9 wrote:
               | > Transformer encoders are not really popular anymore
               | 
               | references, please
        
       | blackeyeblitzar wrote:
       | The thing I still don't understand is how DeepSeek built the base
       | model cheaply, and why their models seem to think they are GPT4
       | when asked. This article says the base model is from their
       | previous paper, but that paper also doesn't make clear what they
       | trained on. The earlier paper is mostly a description of
       | optimization techniques they applied. It does mention pretraining
       | on 14.8T tokens with 2.7M H800 GPU hours to produce the base
       | DeepSeek-V3. But what were those tokens? The paper describes the
       | corpus only in vague ways.
        
         | moralestapia wrote:
         | A friend just sent me a screenshot where he asks DeepSeek if it
         | has an app for Mac and it replies that they have a ChatGPT app
         | from OpenAI, lol.
         | 
         | I 100% believe they distilled GPT-4, hence the low "training"
         | cost.
        
           | Philpax wrote:
           | Er, how would that reduce the cost? You still need to train
           | the model, which is the expensive bit.
           | 
           | Also, the base model for V3 and the only-RL-tuned R1-Zero are
           | available, and they behave like base models, which seems
           | unlikely if they used data from OpenAI as their primary data
           | source.
           | 
           | It's much more likely that they've consumed the background
           | radiation of the web, where OpenAI contamination is dominant.
        
             | Salgat wrote:
             | Hypothetical question: is the chinese government capable of
             | exploiting chatgpt to get around the query limit? For
             | example, making queries through compromised devices or even
             | snooping local traffic on devices? Let's face it, these
             | models are closely alligned with China's national security
             | so it's not a farfetched question to ask.
        
           | kenjackson wrote:
           | They fixed that. Now it replies: "Hi! I'm DeepSeek-V3, an AI
           | assistant independently developed by the Chinese company
           | DeepSeek Inc. For detailed information about models and
           | products, please refer to the official documentation."
        
             | moralestapia wrote:
             | ???
             | 
             | I just did and it told me about ChatGPT and OpenAI.
             | 
             | Are you affiliated with them, btw?
        
               | kenjackson wrote:
               | No. I'm not affiliated. Maybe you're on a node that
               | doesn't have whatever change they might have made.
        
           | blackeyeblitzar wrote:
           | How does such a distillation work in theory? They don't have
           | weights from OpenAI's models, and can only call their APIs,
           | right? So how can they actually build off of it?
        
             | moralestapia wrote:
             | Like RLHF but the HF part is GPT4 instead.
        
               | KarraAI wrote:
               | How do you ensure the student model learns robust
               | generalizations rather than just surface-level mimicry?
        
               | moralestapia wrote:
               | No idea as I don't work on that, but my guess would be
               | that the higher the 'n' the more model A approaches model
               | B.
        
           | nullc wrote:
           | You can't distill from GPT-4 because Open AI conceals the
           | probabilities (and has for a couple years now-- since before
           | gpt4), presumably to prevent that. You can fine tune against
           | output though. I might guess that they used something like
           | openorca or some other public data set that includes gpt4
           | output as part of their initial fine tuning.
        
         | moritonal wrote:
         | I imagine it's a mix of either using ChatGPT as an the oracle
         | to get training data. Or, it's the radiocarbon issue where the
         | Internet has so much info on ChatGPT other models now get
         | confused.
        
         | llm_nerd wrote:
         | Various other models also think they're ChatGPT or built by
         | OpenAI, or at least those are the highest probability tokens
         | when talking about an AI model or an AI company because of the
         | massive prevalence in training data (the internet). It isn't
         | the big reveal that it is often being held to be.
         | 
         | Add that training off of ChatGPT wouldn't reduce their training
         | costs at all, but would actually increase their training costs.
         | Literally all of the same training difficulty, but then add
         | paying OpenAI for an enormous number of API calls. Not really
         | seeing the win.
         | 
         | >The paper describes the corpus only in vague ways.
         | 
         | Anyone who runs a public website has logs absolutely _filled_
         | by a seemingly infinite number of information aggregators. Just
         | like everyone else they scraped the entire internet, pulled in
         | all of Wikipedia, etc. Probably lots of pirate books, movie
         | transcripts, etc.
         | 
         | The fact that training could be done more effectively is
         | something that intuitively makes absolute sense to everyone in
         | the field, but we just didn't make that leap. Similar to how a
         | human isn't trained to recognize digits by training on 60,000
         | training digits then suddenly failing if a real world digit is
         | slightly rotated or morphed in some way, we are making these
         | improvements to content ingestion.
        
       | whoistraitor wrote:
       | It's remarkable we've hit a threshold where so much can be done
       | with synthetic data. The reasoning race seems an utterly solvable
       | problem now (thanks mostly to the verifiability of results). I
       | guess the challenge then becomes non-reasoning domains, where
       | qualitative and truly creative results are desired.
        
         | kenjackson wrote:
         | It seems like we need an evaluation model for creativity. I'm
         | curious, is there research on this -- for example, can one
         | score a random painting and output how creative/good a given
         | population is likely to find it?
        
           | virgildotcodes wrote:
           | How do you account for the impact of culture/lived experience
           | of the specific population viewing the painting? Intuitively
           | it seems like that would be the biggest factor, rather than
           | the objective attributes of the painting, no?
        
             | bentograd wrote:
             | All art is subjective. Any attempt to "verify" a piece of
             | art would be entirely dependent on cultural and personal
             | sensitivities. Art isn't a math problem with a solution.
        
               | baq wrote:
               | But you can dissect it into concepts and see if it is
               | something truly new to the model - if the output contains
               | things which aren't there in the weights, you have a nice
               | specimen to study and, crucially, a recipe to get a bunch
               | of matrices to output untrained things.
        
               | cubefox wrote:
               | This is like saying: All cooks are equally good, even the
               | most disgusting slop (e.g. water/flour soup) isn't any
               | better than a dish from a cook with several Michelin
               | stars. Of course the latter is better. And if it is
               | better, it is objectively better. Even if 0.001% of
               | people prefer flour soup.
        
             | creer wrote:
             | > culture/lived experience of the specific population
             | viewing the painting
             | 
             | Isn't this lived experience baked into LLM language bases?
             | It's certainly very hard to target all possible populations
             | at once. And art doesn't need that, doesn't do that. Only
             | rare marketing sometimes attempts to do that and only in
             | very limited ways, such as a brand name acceptable all over
             | the world.
        
           | esafak wrote:
           | You can train a supervised model, taking into account the
           | properties of the rater as well as the artwork, and tease out
           | the factors that make it rated so.
        
             | creer wrote:
             | You can probably cluster raters and the artwork they rate
             | highly - but probably not in large quantities? -- Which
             | might be the case also with raters being willing to tell
             | you why - and how! most love to do that - but also not in
             | very large quantities. With the added issues that the
             | raters' own opinion of why they love or hate something is
             | likely not to be entirely true and self-understanding.
             | 
             | You could use a larger corpus, like auction house files and
             | art magazines. But then you are confounding for celebrity -
             | a large ingredient in art prices.
        
           | baq wrote:
           | There are two kinds of creativity at play here. One is
           | mashing together combinations of learned things - it's kinda
           | like shuffling a deck of cards where basically every shuffle
           | gets you a deck that has never been seen and won't be seen
           | again, but it's still the same 52 cards every time. The other
           | kind is going outside of the box and inventing truly new,
           | unseen/untrained concepts. This one is hard, but I don't
           | think it's impossible - the <think> slop stirring the learned
           | concepts with a bit of randomness should make progress here.
        
             | kenjackson wrote:
             | A new "AI challenge" -- can an AI make a hit movie (even if
             | just for Netflix) in each of Documentary, Action, Thriller,
             | Comedy, and Drama genres. This isn't art like the "Mona
             | Lisa", but more like the ability to make "art" that has
             | appeal to some level of the public. I think if an AI can do
             | that, I'll be pretty impressed.
             | 
             | The prompt: "Create a feature length [Action/Comedy/etc...]
             | film that can borrow elements from existing films, but
             | would generally not be considered a copy of any given
             | film."
        
           | creer wrote:
           | > can one score a random painting
           | 
           | You can get very mechanical in scoring an image. Ask any art
           | student. If you want to or if your instructor or audience
           | wants to. For example "fits rule of thirds?" yes is a point
           | to common attraction, no is a point to unexpected at the risk
           | of outsider-ness. You can do that in color, composition,
           | recognizing objects and fitting that to memes or associations
           | or non-associations. Too many points in "unexpected" is meta
           | points in "unpleasant chaos" and so a strong downgrade in
           | common attraction. You can match all this to images in the
           | library (see how copyright or song recognition operates in
           | the music category) and get out of that some kind of
           | familiarity vs edge score (where too much edge goes against
           | common attraction.)
           | 
           | I would expect you could get better than most humans at
           | recognizing shapes in an image and drawing associations from
           | that. Such associations are a plus in unexpected / surprise
           | if they are rare in the culture or a plus in common
           | attraction is they are common.
           | 
           | After that, to be cynic about it, you can randomize and
           | second guess yourself so your audience doesn't catch on the
           | 1st level mimicry.
           | 
           | Creativity is not normally used as an absolute with a unique
           | measure. It's not "length". And you only need to please part
           | of the audience to be successful - sometimes a very small
           | part, some of which loves surprise and some hates it, etc.
           | Someone elsewhere objected on the grounds that creativity or
           | attractiveness is culture based - yeah so? if you were to
           | please much of just one whole culture, you would have an
           | insane hit on your hands.
           | 
           | Sounds feasible to me.
        
         | researchers wrote:
         | Tuning for qualitative outcomes is pretty much solved via
         | RLHF/DPO (what this post calls "preference tuning"). Right?
        
         | creer wrote:
         | It's still reasonning based on pattern matching, which should
         | go only so far. But "only so far" could be plenty for lots of
         | applications.
        
       | raphaelj wrote:
       | Do we know which changes made DeepSeek V3 so much faster and
       | better at training than other models? DeepSeek R1's performances
       | seem to be highly related to V3 being a very good model to start
       | with.
       | 
       | I went through the paper and I understood they made these
       | improvements compared to "regular" MoE models:
       | 
       | 1. Latent Multi-head Attention. If I understand correctly, they
       | were able to do some caching on the attention computation. This
       | one is still a little bit confusing to me;
       | 
       | 2. New MoE architecture with one shared expert and a large number
       | of small routed experts (256 total, but 8 in use in the same
       | token inference). This was already used in DeepSeek v2;
       | 
       | 3. Better load balancing of the training of experts. During
       | training, they add bias or "bonus" value to experts that are less
       | used, to make them more likely to be selected in the future
       | training steps;
       | 
       | 4. They added a few smaller transformer layers to predict not
       | only the first next token, but a few additional tokens. Their
       | training error/loss function then uses all these predicted tokens
       | as input, not only the first one. This is supposed to improve the
       | transformer capabilities in predicting sequences of tokens;
       | 
       | 5. They are using FP8 instead of FP16 when it does not impact
       | accuracy.
       | 
       | It's not clear to me which changes are the most important, but my
       | guess would be that 4) is a critical improvement.
       | 
       | 1), 2), 3) and 5) could explain why their model trains faster by
       | some small factor (+/- 2x), but neither the 10x advertised boost
       | nor how is performs greatly better than models with way more
       | activated parameters (e.g. llama 3).
        
         | alecco wrote:
         | They also did bandwidth scaling to handle work around the
         | nerfed H800 interconnects.
         | 
         | > efficient cross-node all-to-all communication kernels to
         | fully utilize IB and NVLink bandwidths
         | 
         | > The key idea of DualPipe is to overlap the computation and
         | communication within a pair of individual forward and backward
         | chunks. To be specific, we divide each chunk into four
         | components: attention, all-to-all dispatch, MLP, and all-to-all
         | combine. Specially, for a backward chunk, both attention and
         | MLP are further split into two parts, backward for input and
         | backward for weights, like in ZeroBubble (Qi et al., 2023b). In
         | addition, we have a PP communication component.
         | 
         | (I know some of those words)
         | 
         | https://arxiv.org/html/2412.19437v1
        
         | whoisburbansky wrote:
         | The key idea of Latent MHA is that "regular" multi-headed
         | attention needs you to keep a bunch of giant key-value (KV)
         | matrices around in memory to do inference. The "Latent" part
         | just means that DeepSeek takes the `n` KV matrices in a given
         | n-headed attention block and replaces them with a lower-rank
         | approximation (think of this as compressing the matrices), so
         | that they take up less VRAM in a GPU at the cost of a little
         | extra compute and a little lost accuracy. So not caching,
         | strictly speaking, but weight compression to trade compute off
         | for better memory usage, which is good because the KV matrices
         | are one of the more expensive part of this transformer
         | architecture. MoE addresses the other expensive part (the
         | fully-connected layers) by making it so only a subset of the
         | fully-connected layers are active at any given forward pass.
        
           | whoisburbansky wrote:
           | https://planetbanatt.net/articles/mla.html this is a great
           | overview of how MLA works.
        
         | sinuhe69 wrote:
         | I think the fact that they used synthetic/distilled high-
         | quality data from GPT4-o output to train in the style of Phi
         | models are of significance as well.
        
       | alecco wrote:
       | How is this very high signal vs noise post out of the front page
       | in 2hs?
       | 
       | Are people so upset with the stock market crash that they are
       | flagging it?
        
         | jasonjmcghee wrote:
         | Maybe too much of the same topic? "How R1 was trained" also
         | seemed to quickly fall off. But the big arxiv paper with 1000+
         | upvotes stuck around a while.
        
           | htk wrote:
           | Spot on. I've read the very accessible paper and it's better
           | than any of the how-to's written elsewhere. Nothing against
           | good content being written, but the source material is
           | already pretty good.
        
         | khazhoux wrote:
         | dang has provided an answer for how the algorithm works
         | whenever I've asked similar question.
         | 
         | But I still don't get it. 6 hours + 170 points and it's on
         | third page. Meanwhile second page has "Null Byte on Steroids"
         | at 12 hours + 20 points. ??
        
       | youssefabdelm wrote:
       | The "illustrated"... He needs to read up on Tufte or Bret Victor
       | or something, these are just diagrams with text inside of boxes.
        
       | 8n4vidtmkvmk wrote:
       | > This is a large number of long chain-of-thought reasoning
       | examples (600,000 of them). These are very hard to come by and
       | very expensive to label with humans at this scale. Which is why
       | the process to create them is the second special thing to
       | highlight
       | 
       | I didn't know the reasonings were part of the training data. I
       | thought we basically just told the LLM to "explain its thinking"
       | or something as an intermediate step, but the fact that the
       | 'thinking' is part of the training step makes more sense and I
       | can see how this improves things in a non-trivial way.
       | 
       | Still not sure if using word tokens as the intermediate
       | "thinking" is the correct or optimal way of doing things, but I
       | don't know. Maybe after everything is compressed into latent
       | space it's essentially the same stuff.
        
       | QuadrupleA wrote:
       | Am I the only one not that impressed with Deepseek R1? Its
       | "thinking" seems full of the usual LLM blindsides, and ultimately
       | generating more of it then summarizing doesn't seem to overcome
       | any real limits.
       | 
       | It's like making mortgage backed securities out of bad mortgages,
       | you never really overcome the badness of the underlying loans, no
       | matter how many layers you pile on top
       | 
       | I haven't used or studied DeepSeek R1 (or o1) in exhaustive
       | depth, but I guess I'm just not understanding the level of
       | breathless hype right now.
        
         | itissid wrote:
         | It is leaps and bounds better than LLMs. For one you are doing
         | RL which is classic AI like tuning that optimizes a reward
         | function with nice qualities- it's the same stuff used to train
         | chess games and Go by showing them the actual moves.
         | 
         | LLMs pre o1 and deepseek R1 were RHLF tuned which is like if
         | you trained a LM how to play chess by showing people two boards
         | and doing a vibe check on which "looks" better.
         | 
         | Think of it this way say you were dropped in a maze that you
         | had to solve but you could do only one of two things:
         | 
         | 1. Look at two random moves from your start position and
         | selected which one looked better to get out.
         | 
         | 2. Made a series of moves and then backtracked, then use a
         | quantitative function to exploit the best path.
         | 
         | The latter is what R1 does and it chooses optimal and more
         | certain path to success.
         | 
         | Apply this to math and coding tokens and you have a competitive
         | LLM.
        
           | itissid wrote:
           | I am using the 32Gb distilled model on my local 3090 with
           | Continue in VSCode. It beats everything out of the water.
        
             | dontwearitout wrote:
             | How many tokens/s do you get on a 3090? With the extra
             | tokens for the internal monologue, is it still performant
             | enough for smooth VSCode integration?
        
             | MarcelOlsz wrote:
             | Any idea how to use a cloud hosted version with cursor?
        
         | throwup238 wrote:
         | You're not the only one. It's not as impressive at coding
         | compared to O1 as people make it out to be and it's explicitly
         | spelled out in DeepSeek's R1 paper that they had trouble with
         | improving over DeepSeek-V3:
         | 
         |  _> Software Engineering Tasks: Due to the long evaluation
         | times, which impact the efficiency of the RL process, large-
         | scale RL has not been applied extensively in software
         | engineering tasks. As a result, DeepSeek-R1 has not
         | demonstrated a huge improvement over DeepSeek-V3 on software
         | engineering benchmarks. Future versions will address this by
         | implementing rejection sampling on software engineering data or
         | incorporating asynchronous evaluations during the RL process to
         | improve efficiency._ (last bullet point of page 16, which is
         | the last page of the paper before the bibliography - hmm...)
         | [1]
         | 
         | It does even worse on my real world coding problems than the
         | benchmarks would suggest. Some of my tests: write a Qt
         | QAbstractListModel in C++ that parses markdown into block using
         | a C/C++ MD parsing library, write Rust cxx-qt bindings for
         | QTextDocument (all docs included in context), write a window
         | switching script for Wayland with an alternative to wmctrl,
         | etc. I also asked it some geochemistry questions that I had
         | trialed O1 with previously, and the reasoning had a lot
         | hallucinations. The answers were suboptimal to say the least.
         | 
         | Having access to the thought process in <think></think> tags is
         | cool but it just reveals how underdeveloped the actual
         | reasoning is compared to whatever o1 is doing. I've had it get
         | stuck on silly things like whether a C++ library for markdown
         | exists because I underspecified that I'm okay with header only
         | C libs. O1 has faired much better on my qualitative test suite
         | - even more so when using it via API with a "high"
         | reasoning_effort.
         | 
         | With all of the hype over the last few days, it feels like I'm
         | taking crazy pills (loving all the action though).
         | 
         | [1] https://arxiv.org/abs/2501.12948
        
         | lm28469 wrote:
         | I've been using deepseek for a while, I never paid for chat gpt
         | or any other services.
         | 
         | The fact that r1 is now free and unlimited VS chat gpt 200$ a
         | month subscription is impressive enough for me. If the
         | development cost is anywhere close to what they advertise
         | publicly it's even more impressive
         | 
         | It's as good or better than chat gpt free, gemini free, &c. and
         | that's all I care about
        
           | redcobra762 wrote:
           | Why is cost your only concern? You don't care at all what
           | data the model was trained on? The motivations of the people
           | who trained it? I mean maybe the importance of those things
           | isn't super high to you, but "don't care"?
           | 
           | How odd. Most Westerners I know would care...
        
             | lm28469 wrote:
             | Ah yes, because chat gpt is ethical AI lmao
             | 
             | I'm European, seeing the clowns in charge of US tech
             | companies I'm equally happy using Chinese tools, especially
             | if they're open source.
             | 
             | Stop twisting the debate as if it was "good US" vs "evil
             | China", I have no horses in this race
        
               | redcobra762 wrote:
               | That's fine, it's just not what Westerners generally say,
               | in my experience. Not that distaste for Chinese
               | authoritarianism is inherently pro Western, just that
               | Westerners I know tend to have an ideological issue with
               | the substantially higher levels of censorship and
               | oppression that takes place in China.
               | 
               | It's surprising to find people who genuinely don't care
               | about any of that, is all.
        
               | lm28469 wrote:
               | "westerners" as if it was a monolithic block. I envy the
               | simplicity of seeing life in such terms, China bad, US
               | good, it must be very relaxing
               | 
               | You don't have any concerns with the US government, the
               | nepotism, the corruption, the conflict of interests, the
               | insider trading, the massive concentration of wealth and
               | power in tech, the insane lobbyism, PRISM, &c.
               | 
               | It's surprising to find people who can so clearly see how
               | China is bad but are completely oblivious to their own
               | problems. It's not a football game, you don't have to
               | chose a side and be a boot licker for eternity
        
               | redcobra762 wrote:
               | I didn't say China bad, I said Westerners typically find
               | the levels of oppression and censorship present in China
               | to be of concern. I actually gave zero of my own
               | judgement on China at all.
               | 
               | If you really think the Good vs. Evil narrative is wrong,
               | why would you immediately go towards unrelated generic
               | issues the West has? A neutral party would be more likely
               | to acknowledge the problems with both sides, not
               | reflexively try to change the subject!
               | 
               | Then again you didn't claim to be a neutral party, did
               | you?
        
               | lm28469 wrote:
               | Well yeah, both sides are fucked so I'll use the free
               | tool and not the $200/month tool, it really isn't rocket
               | science
               | 
               | Even if deepseek is a chinese communist party evil trick
               | what do they get ? My shitty code ? Big deal, at least
               | I'm not down $200 a month, which is half of my rent
        
               | lukan wrote:
               | If they would establish deepseek in their authorized
               | version worldwide - it means they would establish their
               | worldview worldwide.
               | 
               | Students asking for help with homework will get the
               | chinese approved version. Any housewife asking for
               | recommendation who to vote and why, etc.
               | 
               | But ... deepseek seems open source and the local version
               | not restricted (as it likely was trained on ChatGPT in
               | the first place). So I also refuse to refuse deepseek
               | because it comes from china. I see it as more competition
               | that hopefully will help with establishing good open
               | source models in control of no single political
               | organisation.
        
               | lm28469 wrote:
               | > Students asking for help with homework will get the
               | chinese approved version. Any housewife asking for
               | recommendation who to vote and why, etc.
               | 
               | Well too bad for them, using a pencil to cut bread sucks
               | too, that's why we don't do it. Meanwhile it's converting
               | my json to xml and pissing code so I don't have to piss
               | it myself
        
               | lukan wrote:
               | Too bad for you, if the majority of them will vote
               | someone not to your liking.
        
               | lm28469 wrote:
               | That's why I'm investing in land and ammos instead of
               | cryptos and nvidia stocks
        
               | lukan wrote:
               | I can see the motive, but I rather invest in a world
               | where I won't end up in a fortified home shooting hungry
               | scavengers.
        
               | lm28469 wrote:
               | I'm investing in my local community, it's closer to
               | chicken coop vs rabbit cages than chatgpt vs deepseek,
               | and sadly the war isn't even 1000km away as we speak
        
               | lukan wrote:
               | A bit closer for me, but how will the local community
               | help, if the war is coming?
               | 
               | I considered switzerland for this reason. (Also I like
               | mountains)
        
               | acka wrote:
               | Off topic: why the hell are you converting json to xml? I
               | would rather convert json to yaml instead, a step
               | forwards instead of backwards in evolution, but that is
               | my opinion.
        
               | lukan wrote:
               | Working with legacy backward code?
        
               | cratermoon wrote:
               | > Students asking for help with homework will get the
               | chinese approved version. Any housewife asking for
               | recommendation who to vote and why, etc.
               | 
               | How is this any worse than where the US is now, getting
               | the Musk/Zuckerberg/Bezos/Trump approved answers, if any
               | answer can be gotten at all, after the current occupant
               | of the oval office has silenced all federal agencies.
        
               | redcobra762 wrote:
               | Western trained models will tell you politically
               | inconvenient things, and if they won't, that info is
               | otherwise freely available. Neither is the case for
               | Chinese trained models.
        
               | cratermoon wrote:
               | > Western trained models will tell you politically
               | inconvenient things,
               | 
               | Pull the other leg. Do you know who David Mayer is? Hint:
               | https://archive.ph/iI5xC
        
               | redcobra762 wrote:
               | It's a random guy. He's also not censored by OpenAI
               | models:
               | 
               | > Who is David Mayer
               | 
               | ![Quien es David Mayer de Rothschild, el millonario al
               | que relacionan ...](https://tse3.mm.bing.net/th?id=OIP.CC
               | KTcyWe80YvrRbkkxY_AAHaK...) The name "David Mayer" is
               | associated with several individuals across various
               | fields:
               | 
               | 1. *David Mayer de Rothschild*: Born in 1978, he is a
               | British environmentalist, adventurer, and member of the
               | Rothschild family. He is known for his environmental
               | advocacy and expeditions, including the "Plastiki"
               | project, where he sailed across the Pacific Ocean on a
               | boat made from recycled plastic bottles to raise
               | awareness about plastic pollution. ([en.wikipedia.org](ht
               | tps://en.wikipedia.org/wiki/David_Mayer_de_Rothschild?utm
               | _...))
               | 
               | 2. *David M. Mayer*: A professor of Management and
               | Organizations at the University of Michigan, specializing
               | in behavioral ethics. ([scholar.google.com](https://schol
               | ar.google.com/citations?hl=en&user=c2Zunb8AAAAJ...))
               | 
               | 3. *David R. Mayer*: An American politician born in 1967,
               | he has served as the mayor of Gloucester Township, New
               | Jersey, and was a member of the New Jersey General
               | Assembly. ([en.wikipedia.org](https://en.wikipedia.org/wi
               | ki/David_R._Mayer?utm_source=chat...))
               | 
               | 4. *David Delaney Mayer*: Born in 1992, he is an American
               | documentary filmmaker and social entrepreneur, known for
               | projects like the PBS series "Food Town" and co-founding
               | DreamxAmerica, an initiative supporting immigrant
               | entrepreneurs. ([en.wikipedia.org](https://en.wikipedia.o
               | rg/wiki/David_Delaney_Mayer?utm_source...))
               | 
               | 5. *David Mayer (Historian)*: An American-British theatre
               | historian (1928-2023), he was an emeritus professor at
               | the University of Manchester, recognized for his work on
               | 19th-century drama and the Victorian stage. ([en.wikipedi
               | a.org](https://en.wikipedia.org/wiki/David_Mayer_%28histo
               | rian%29?ut...))
               | 
               | 6. *Akhmed Chatayev (Alias: David Mayer)*: A Chechen
               | militant (1980-2017) who used the alias "David Mayer."
               | This alias led to a case of mistaken identity affecting
               | the historian David Mayer. ([en.wikipedia.org](https://en
               | .wikipedia.org/wiki/David_Mayer_%28historian%29?ut...))
               | 
               | If you have a specific "David Mayer" in mind, please
               | provide more context or details, and I can offer more
               | targeted information.
        
               | antifa wrote:
               | This is virtue signaling and concern trolling until you
               | post evidence that DeepSeek is doing worse than what
               | ChatGPT does on Israel vs Palestine.
        
               | redcobra762 wrote:
               | Who am I signaling my virtue to if I'm posting
               | anonymously?
        
               | raxxor wrote:
               | The CEO did gave a statement about their motivation.
               | Could be a lie, but he delivered and it is also vastly
               | more sensible that what we often hear from other
               | companies. Google and Meta are an exception for this
               | space though.
               | 
               | Also, because not only the weights, but also the data is
               | open, any propaganda can be identified and corrected.
               | This is not the case for other models and what we have
               | seen from Gemini, there certainly are "adaptations". I
               | don't think Google had ill intent here, but this would
               | fit what some would classify as propaganda.
        
               | mistrial9 wrote:
               | this directly blends people with a government.. Western
               | people is not equal to the USA govt. Here in California,
               | three decades of China human rights history is very clear
               | to a lot of people. In fact, many Chinese speaking people
               | on the Americas west coast, left China for specific
               | reasons, too.
               | 
               | Utilitarian money-oriented self-servers definitely have
               | less "care" about these things? EU or UK or wherever
        
               | gooosle wrote:
               | It's funny you say that cause chatgpt, Gemini, etc have
               | way more censorship built in than deepseek.
        
             | richardw wrote:
             | I think there's a real chance of changes in soft-power
             | dynamics given recent events. When allies feel their
             | partner and neighbour is behaving solely in its own
             | interests, it changes trust relationships.
             | 
             | Trump, tech, US-first focus with less interest in
             | collaboration. TikTok. Ukraine and discussions around
             | defence budgets. Reducing focus on EV's. Tariffs. Iceland.
             | Gulf of America.
             | 
             | Immediate choices might be fear-based but it's smart to
             | look for other partners as trust erodes.
             | 
             | An extreme side effect might be countries who felt safe
             | under the US arms umbrella, needing to arm themselves. Are
             | we absolutely sure that's what we want? Does that include
             | nukes?
        
             | exe34 wrote:
             | Chinese models are censored, while Western models are
             | _aligned_. it's a very important distinction.
             | 
             | personally I imagine in the future I'll use a mock UN panel
             | of LLMs to advise me, and avoid any one nation/political
             | party's influence, if I ever get to the point of delegating
             | much of my thinking to the machine.
             | 
             | Most westerners wouldn't want their land threatened by
             | their allies either, but we don't live in a perfect world.
        
               | premysl wrote:
               | Talk to R1 for a while, and you'll notice that it's both
               | censored and aligned.
               | 
               | I think the most free-minded large models might be the
               | Groks, but just slightly, as they have different biases.
               | In sum, there's strength in diversity.
        
               | lm28469 wrote:
               | > Western models are _aligned_
               | 
               | Aligned to what ? By who ? Did you vote for any of these
               | people ? Did anyone ask you your opinion ? Can I see what
               | topics are aligned ? Can I see how much they are aligned
               | ? Does the alignment change depending on who's the
               | current US president ?
        
               | exe34 wrote:
               | that was my point, precisely.
        
               | cratermoon wrote:
               | > a very important distinction.
               | 
               | A distinction without a difference.
               | 
               | As used by the Musk/Zuck/Bezos crowd, "aligned" is a
               | weasel word for "parroting the TESCREAL world view"
        
               | exe34 wrote:
               | Indeed.
        
             | troyvit wrote:
             | > You don't care at all what data the model was trained on?
             | The motivations of the people who trained it?
             | 
             | Working for a news org whose data was used without our
             | consent to give ChatGPT a leg up, yeah I'd care a lot about
             | it. The bottom line though is that nobody is innocent in
             | that regard.
             | 
             | Would a self-respecting organization go all in on Deepskeek
             | R1 without a security audit and a ton of competitive
             | testing against other models? I doubt it. Same way they
             | shouldn't just give all their employees OnePlus or Huwei
             | phones.
             | 
             | I like how the Meta guy put it. This isn't China beating
             | the US in AI, it's "open source" (however these guys define
             | it) beating closed models.
             | 
             | OpenAI broke its promise with the world, which anybody who
             | compares their name with their product can tell you. If
             | open models put a dent in their hegemony it's only good for
             | the rest of the industry.
        
               | redcobra762 wrote:
               | So it genuinely does not matter to you how completely
               | irrelevant US copyright law is to China, when using a
               | Chinese LLM?
               | 
               | That's such a weirdly specific application of a moral
               | principle, it's hard to believe.
        
               | troyvit wrote:
               | It doesn't matter to you how completely irrelevant global
               | copyright law is to OpenAI? [1]
               | 
               | [1] https://futurism.com/the-byte/openai-copyrighted-
               | material-pa...
        
               | redcobra762 wrote:
               | No, but then again I didn't claim copyright law mattered
               | to me at all, you did.
        
               | troyvit wrote:
               | I also said:
               | 
               | > The bottom line though is that nobody is innocent in
               | that regard.
               | 
               | In other words, that bird has flown and it isn't a valid
               | reason to choose one model over another.
        
               | redcobra762 wrote:
               | Except _you_ said it was worth caring about! I agree with
               | what you wrote here, but I thought we were talking about
               | this because you had concerns about OpenAI 's disrespect
               | for copyright.
        
               | CamperBob2 wrote:
               | It doesn't matter to the Chinese, and that's all that
               | matters.
               | 
               | To the extent that training a model with copyrighted
               | material infringes copyright, copyright law has to
               | change. There _is_ no other point of view. Disagreement
               | means forfeiting a very important game before it even
               | begins.
        
         | assimpleaspossi wrote:
         | fwiw, National Public Radio (NPR) news in the USA said that AI
         | experts stated it was "almost as good" as other current
         | offerings like chatGPT and Gemini. That its real advantage is
         | the low cost of providing the information. However, this is
         | only a claim made by the company without any proof.
        
         | TrackerFF wrote:
         | What's there not to understand?
         | 
         | If it matches the latest GPT O-N model in performance - or is
         | just close, even, at a fraction of the compute (50x less?) and
         | it is free, then that's huge news.
         | 
         | They just upended the current LLM/AI/ML dominance, or at least
         | the perceived dominance. Billions and billions have been pumped
         | into the race, where investors are betting on the winner - and
         | here comes a Chinese hedge fund side-project on shoestring
         | budget, matching those billion dollar behemoths. And they'll
         | continue to release their work.
         | 
         | They just made the OpenAI et. al. secret sauce a lot less
         | valuable.
        
           | infecto wrote:
           | In my tests it does not come close to O1-Pro. Still huge news
           | but it does not quite make it.
        
             | rallyforthesun wrote:
             | The results i did get from deepseek-r1 on their webpage did
             | not match the results i did get from o1-pro. I did ask it
             | go to a github repo, find the part where the logic of the
             | "export" button is and explain why it doesn't work (the
             | whole logic is actually missing, won't work at all). O1 pro
             | did get it right in the first try while deepseek r1 was
             | heavily hallucinating. Maybe i am using the wrong model?
        
               | throwup238 wrote:
               | No, you're not. They explicitly mention in the R1 paper
               | (in the last paragraph before the bibliography) that R1
               | isn't a "huge" improvement over DeepSeek-V3 in coding -
               | where "huge" is an academic weasel word.
               | 
               | It's just a lot of hype. In my coding tests it
               | significantly underperforms o1 (haven't tried o1-pro),
               | often getting stuck in a reasoning loop because I
               | underspecified something (that I don't have to with o1).
        
               | infecto wrote:
               | Same anecdotal experience. Its definitely an improvement
               | and they have made operational improvements at runtime
               | but I am still concerned they are have over fit for the
               | tests.
        
           | thechao wrote:
           | New theory: this is a short-long play by the fund. They
           | shorted NV, now they're hoovering up stock. In the process of
           | making their billions from a small 50$mm investment!
        
             | zombiwoof wrote:
             | Bullseye
        
               | thechao wrote:
               | Damn Matt Levine to hell! His latest newsletter has an
               | entire section devoted to the entire topic!
        
         | creer wrote:
         | The brute force approach is the expensive one ("probably, so
         | far anyway" add these 4 words everywhere) - and impossible to
         | make "always correct" aka "usual blindsides". They seem to be
         | trying a bunch of specialized training ideas here and there in
         | the system - just like a Mixture of Experts does, in different
         | places than was obvious so far, and with an eye toward
         | reasonning. In particular trying to build a reasonning-oriented
         | training base from minimal seed.
         | 
         | It's still not going to give an "always correct" result but we
         | are nowhere near the point where that's needed. We are only at
         | the point where a new idea can get you a percentage point
         | further in benchmarks. Some fundamental limits were baked into
         | the previous assumptions - easy to get past by leaving these
         | assumptions behind.
        
         | dkjaudyeqooe wrote:
         | As with most early tech of a particular category, it's not the
         | current capabilities that are the point but the direction of
         | travel.
         | 
         | DeepSeek has upended the conventional wisdom about model
         | performance with respect to training, and it's a shock to the
         | system. It demonstrates something that has become obvious: you
         | don't need massive scale or funding to innovate and have
         | impact, you just need good ideas.
        
         | HarHarVeryFunny wrote:
         | It's the cost comparison with O1, both to train and run (per
         | their pricing), that is causing most of the shock, perhaps as
         | well as the fact that it's a GPU poor Chinese company that has
         | caught up with O1, not a US one (Anthropic, Google, Meta, X.ai,
         | Microsoft). The fact that it's open weights and training is
         | fairly detailed in the paper they released is also significant.
         | 
         | The best comparison for R1 is O1, but given different training
         | data, hard to compare outside of benchmarks. At the moment
         | these "reasoning models" are not necessarily the best thing to
         | use for non-reasoning tasks, but Anthropic have recently
         | indicated that they expect to release models that are more
         | universal.
        
         | vineyardmike wrote:
         | I get the hype, even if I don't necessarily agree with it.
         | TLDR: it's technically impressive in training infrastructure
         | and a geopolitical surprise.
         | 
         | As others commenters said, it compares favorably against
         | American Flagship models in benchmarks. This is geopolitically
         | interesting since the model is Chinese, and subject to trade
         | restrictions and America thinks of itself as the world's best
         | model builders.
         | 
         | What makes it interesting technically, is the trade
         | restrictions required them to focus on efficiency and reusing
         | old hardware. This made it wildly cheaper to run the training.
         | The They make a ton of very low-level optimizations to make
         | training efficient. This is impressive engineering, and shows
         | how a small amount of effort can free up a nvidia-lock-in. They
         | had to totally bypass a lot of CUDA libraries and drop down
         | even lower to control the GPUs. Nvidia has been capturing a
         | huge fraction of industry wide AI spend, and surely no one
         | wants that. This is, IMO, the actual part to watch. I suspect
         | it'll drive a new wave of efficiency-focused LLM tools which
         | will also unlock competitors GPUs.
         | 
         | They also had some novel-for-LLMS training techniques but it's
         | suspected that the big AI companies elsewhere are doing it now
         | too, but not disclosed. (Mostly reinforcement learning).
         | 
         | What I think is hype, meanwhile, is the actual benchmarks. Most
         | models are trained on their competitors output. This is just a
         | reality of the industry. Especially non-flagships being trained
         | against flagships data. DeepSeek was almost certainly trained
         | against OpenAI models, so it makes sense that it would approach
         | the output quality. That's very different from being capable of
         | outperforming or "taking the lead" or whatever. We'll need to
         | wait longer and see how the future goes to make that
         | determination. "China" has long had a great history of machine
         | learning tech, so there is no reason to think that it's
         | structurally impossible for a Chinese organization to be on the
         | leading edge, but it has to happen before we can say it
         | happened.
         | 
         | What is also hype is calling this a "side project" of a
         | financial firm. The firm spun it out as a dedicated company.
         | China cracked down on hedge funds, so the company looked for
         | ways to repurpose the talent and compute infrastructure. This
         | isn't some side project during lunch breaks for a few bored
         | quants.
         | 
         | PS, thinking models are very different in use-case than normal
         | models. It's great for tasks with a "right answer" like math
         | problems. Consider a simple example, in calculus, your teacher
         | made you "show your work", and doing so certainly reduced the
         | likelihood of errors. That's the purpose of a thinking model,
         | and it excels at similar tasks.
        
       ___________________________________________________________________
       (page generated 2025-01-28 23:01 UTC)