[HN Gopher] Exclusive access for LLM companies to largest Chines...
       ___________________________________________________________________
        
       Exclusive access for LLM companies to largest Chinese nonfiction
       book collection
        
       Author : sillysaurusx
       Score  : 206 points
       Date   : 2023-11-05 07:30 UTC (15 hours ago)
        
 (HTM) web link (annas-blog.org)
 (TXT) w3m dump (annas-blog.org)
        
       | sillysaurusx wrote:
       | This collection is 40 times larger than books3. No one seems to
       | care at all about copyright. It's tempting to leave the English
       | speaking world behind and focus on someplace that supports
       | academic work.
       | 
       | Will people call this theft? No one has so far.
       | 
       | In the meantime, simply abandoning European and American
       | knowledge is a realistic option. LLMs will learn in Chinese.
       | 
       | On the other hand, it would suck much more to be thrown in a
       | Chinese prison. So probably don't do this if you plan on visiting
       | China. (If China reacts by fully endorsing this, that would be
       | the uno reverse of the century.)
        
         | squigz wrote:
         | Sharing of information is more important to me personally than
         | business interests
        
         | alex_duf wrote:
         | China may find it attractive to have a model trained on their
         | languages and cultures.
         | 
         | It would send a message of power, and ensure that local tech
         | companies will use a model with their influence rather than
         | western corporate influence.
         | 
         | Not saying they should or would endorse, but it wouldn't be
         | _that_ surprising.
        
         | sillysaurusx wrote:
         | Replying to a deleted comment, because I think this is worth
         | underscoring:
         | 
         | By explicitly supporting this, China could signal to AI
         | researchers worldwide that China is a safe haven. It's the
         | materialization of what everyone said would happen: when
         | copyright enthusiasts shut down AI work in America, companies
         | will simply go elsewhere.
         | 
         | Imagine if the epicenter of AI research shifted away from the
         | US. This may seem unlikely, but lack of data is the biggest
         | problem that open source LLMs face. Mistral had to completely
         | refuse to disclose their training sources just to release a
         | decent model.
         | 
         | Personally, I'm tempted to focus on this full time. It's what
         | I've always wanted to do. I never gave a thought to legal
         | consequences, and with this collection there appears to be
         | none.
        
           | TerrifiedMouse wrote:
           | > China is a safe haven
           | 
           | Heh. It's a safe haven until your LLM says something the CCP
           | doesn't like.
        
             | FartyMcFarter wrote:
             | "Safe"and "haven" are indeed bold worlds when applied to
             | the current Chinese regime...
        
             | Vetch wrote:
             | This issue is overstated. Given the material is in Chinese,
             | it stands to reason the bulk (but of course not all) of it
             | will not be in violation of whatever policies. Furthermore,
             | there have been a series of open-weights Chinese models
             | that give reasonable answers to sensitive (in China)
             | questions. Unless you plan to release something customer
             | facing in China, it's not something to stress about.
             | 
             | The idea that overly burdensome regulations around open
             | source models in the US shifts the global center of mass of
             | LLMs to China is not implausible.
             | 
             | One reason resonant among government officials against
             | opensource LLMs is to keep them from China and Russia. But
             | among the very best embeddings, ~13Bs and perhaps 34B open-
             | weights models are the Chinese ones. The recent
             | DeepSeekCoder also tops EvalPlus:
             | https://evalplus.github.io/leaderboard.html For those
             | concerned about over-training, ChatGPT and GPT4 are almost
             | certainly also contaminated and there have been independent
             | confirmations of DeepSeekCoder's strength.
        
               | rfoo wrote:
               | > Given the material is in Chinese, it stands to reason
               | the bulk (but of course not all) of it will not be in
               | violation of whatever policies.
               | 
               | Especially due to no freedom of the press in China (you
               | need a permit for each published book) they are all
               | already censored.
               | 
               | I still don't think China could be a safe haven though,
               | they haven't shot down open source models not because
               | they are nice, it's just that they are incompetent.
        
               | TerrifiedMouse wrote:
               | > Especially due to no freedom of the press in China (you
               | need a permit for each published book) they are all
               | already censored.
               | 
               | Even in a fully censored environment, you can still
               | accidentally run afoul of state censors.
               | 
               | https://www.theguardian.com/world/2022/jun/06/chinese-
               | star-t...
        
               | corethree wrote:
               | No LLMs will evolve emergent behavior. It will have
               | capabilities beyond the training data including saying
               | shit about china. You just need the to prompt it
               | correctly.
        
           | stavros wrote:
           | While this is true, imagine having to run your entire dataset
           | past the censors, to make sure your LLM doesn't turn out to
           | have unpalatable opinions.
        
             | manquer wrote:
             | Only if the model serves the mainland. Chinese authorities
             | don't care what you serve others I think .
        
             | greesil wrote:
             | In Soviet China, LLM censors you.
             | 
             | Sort of joking, but some people seem to be using the dual
             | LLM approach where the 2nd LLM censors the output of the
             | first. Take the Gandalf game as an example.
             | 
             | And while you don't need an LLM to censor an entire
             | internet at scale I see how it would be a useful tool for
             | the censors in China or anywhere else. "Hey CensorGPT does
             | the following weibo post history make fun of dear leader,
             | taking note of the themes of Winnie the Pooh, touch on any
             | other criticism of the CPC" etc
        
             | lucubratory wrote:
             | On a technical level it's really not that much harder than
             | all of the RLHF companies have to do in the West.
        
               | stavros wrote:
               | I guess it wouldn't be, since it's for the exact same
               | purpose. The west is just censoring different things than
               | the east.
        
               | squigz wrote:
               | They both censor what they don't like
        
           | m3kw9 wrote:
           | Maybe they will need you to give them access to weights like
           | how tech co need to share IPs with them.
        
         | yorwba wrote:
         | Anna's Archive is already blocked by the Great Firewall
         | https://blocky.greatfire.org/detail/535678/https%3A%2F%2Fann...
         | , chance of endorsement is nil.
        
           | sillysaurusx wrote:
           | Is there a way to find out why? It could be anything from
           | "most new websites get blocked by default, or automatically"
           | to "a censor didn't like a specific thing."
           | 
           | I'm not a fan of bowing to censorship (to say the least), but
           | at this point I am even less a fan of being verbally spit on,
           | called a thief, and being included in lawsuits. I'd take the
           | lesser of two evils, since there's no way to even do any work
           | in the first place in America.
           | 
           | (I've been depressed for months that the inevitable outcome
           | of all the LLM lawsuits is that book authors will demand
           | licensing fees, which completely kills open source AI models
           | for all but the most funded institutions. I.e. my original
           | work would've been completely impossible.)
        
             | yorwba wrote:
             | Most websites do not get blocked. When it happens, the
             | censors won't tell you, but if there's a search bar, it's
             | usually easy to find content that explains the block:
             | https://annas-
             | archive.org/search?q=Prisoner+of+the+State+Zha...
        
               | sillysaurusx wrote:
               | Thank you. I have no way of knowing anything about China,
               | so this kind of thing is super helpful. I wouldn't have
               | even known what to search for to trigger the censors.
        
               | rfoo wrote:
               | In principal it's easy to explain - as long as you have
               | uncensored content on your website, it's going to be
               | blocked. For example, HN is blocked.
               | 
               | If it didn't happen, then it's because you are not big
               | enough.
               | 
               | A few cases:
               | 
               | - FB, Instagram, Reddit, Twitter, Youtube are all
               | blocked: user generated content and without outrageous
               | censorship effort (Douyin has 100k employees dedicated to
               | censor and is top cost center, Weibo 50k IIRC) it is very
               | predictable, just assume such things get blocked.
               | 
               | - Google is blocked. They announced in 2010 they didn't
               | want to censor any more.
               | 
               | - A censored version of Bing is not blocked.
               | 
               | - Dropbox is blocked: the same reason.
               | 
               | - Notably, GitHub is NOT blocked. GitHub pages and Gist
               | are blocked. This is because GitHub does censor repos
               | upon government request, plus that when it was once
               | blocked scientist spent effort and managed to reverse the
               | decision.
               | 
               | - WhatsApp and Signal are blocked. SMS from Signal is
               | also blocked for all +86 numbers so people simply can't
               | register.
               | 
               | - iMessage is not blocked. This is, IMO, an outlier. My
               | guess is it's because Apple + no group chat +
               | incompetence.
        
               | error_logic wrote:
               | Didn't Apple agree to store data inside China for their
               | citizens in a separate server system, thus making it
               | subject to scrutiny on a national level? If that has been
               | the case and people think otherwise...well, it shows how
               | hard it is to navigate the abstractions of networked
               | systems.
        
               | rfoo wrote:
               | They did. But iMessage is end-to-end encrypted. As long
               | as you avoid the iCloud caveat (iCloud backups are by
               | default not E2E encrypted) you get an actual uncensored
               | E2EE messaging app in China.
        
         | BiteCode_dev wrote:
         | If the AI is accessible to the rest of the world, it's not a
         | problem: the model can query its knowledge and answer in any
         | language since it already is very good at translating things.
         | 
         | The problem is rather than, given the economic war between the
         | US and China, the chip ban, the taiwan issues and the GFC, we
         | may not have access to this.
        
         | morelisp wrote:
         | > simply abandoning European and American knowledge is a
         | realistic option.
         | 
         | Haha, OK, yeah.
        
       | fithisux wrote:
       | The biggest theft in history with governments as complices
        
         | sillysaurusx wrote:
         | With the most valuable haul of all: not gold or coin, but
         | knowledge. How terrible.
         | 
         | But no, I don't think governments are accomplices. The smartest
         | move would be for China to wait and see how this plays out. It
         | could significantly shift the epicenter of AI work to China.
         | The single biggest limiting factor for open source LLMs is
         | data, and no one in America or Europe wants to be sued.
        
           | fragmede wrote:
           | So fund it with crypto and release it on 4chan.
        
           | pilimi_anna wrote:
           | For what it's worth, we also have millions of previously
           | unreleased (in bulk) books in English, mostly non-fiction,
           | that are available for torrent on our website.
           | 
           | High-speed access available for anyone who can do at-scale
           | text extraction, or who can supply us with new collections.
        
             | sillysaurusx wrote:
             | Anna! I just want to say, I love you. Everything about what
             | you're doing is heroic. Whoever and wherever you are, thank
             | you.
             | 
             | Please focus on your opsec. The more visible you become,
             | the progressively angrier people will get. Don't do
             | anything silly like edit your Wikipedia page from your
             | house.
             | 
             | With that out of the way, someone I know happens to have
             | the original books3 epub files. I think they can be
             | convinced to send them to you. It's only 200,000 books, but
             | that could theoretically grow your collection by 10% or so.
             | I don't know whether that would be helpful to you (you've
             | far surpassed books3 at this point), but if so, let me
             | know.
             | 
             | Given the legal risks, the best course of action for AI
             | companies is probably to ignore English and European books
             | entirely. There is plenty of Chinese data, and the models
             | would learn all the same concepts without exposing anyone
             | to lawsuits.
        
               | pilimi_anna wrote:
               | Yes please, put us in touch by email. Or feel free to
               | email me yourself and we can set up more secure comms
               | from there. Thanks so much for everything you are doing
               | as well!
        
               | stavros wrote:
               | Since you're pretty knowledgeable about these things, I
               | think I should ask here: I've made a fairly simple design
               | for a program based on BitTorrent, that will allow people
               | to "donate" their disk space to organizations like
               | archive.org, Anna's Archive, and anything else that needs
               | data hosted.
               | 
               | Basically, you download a client, say "allocate 2 TB of
               | my disks to whatever archive.org/donate/disk.rss" says,
               | and the server/client combination ensures you download
               | and seed the rarest 2TB of the collection.
               | 
               | This design is also open, in the sense that the server
               | can share the database of torrents it contains, and
               | anyone can use it to fetch any of the files in the
               | dataset from the swarm.
               | 
               | Would something like this be at all useful? I've emailed
               | a few archivists, but I got no response, and the one
               | person I've managed to talk to about this said there have
               | been a few attempts on this, but they always fail for one
               | reason or another.
        
               | sillysaurusx wrote:
               | You are literally building what I've been slowly working
               | towards on my own. This seems like a very good sign.
               | Multiple simultaneous discovery is a common occurrence in
               | the sciences.
               | 
               | The hard part is that those who donate their space have
               | authority over that space. It's the Byzantine fault
               | tolerance problem: imagine if 4chan donated their space,
               | then started serving CSAM instead of the expected data.
               | You can use hashes to verify integrity, but then the
               | question becomes who gets to decide which hashes are ok.
               | And hashing makes it impossible to edit large files,
               | which is a frequent occurrence in LLM work. You're
               | constantly tweaking your datasets and spitting out new
               | blobs.
               | 
               | Direct answer: yes, you're doing good work, and you
               | should keep doing it. I would personally use this for
               | storing books3 transformations.
               | 
               | The other hard part is that you'll want at least some
               | redundancy -- see 6.824 distributed systems, or the GFS
               | paper. It's why I've been implementing Raft and toying
               | with some kind of distributed consensus without a
               | blockchain. (Such consensus is still possible if the
               | researchers were granted authority over what can be
               | stored -- which is the whole reason people are donating
               | their disk space in the first place.)
               | 
               | Another issue is sudden bandwidth loss. Data storage is
               | one part of it. The other half is rapid transfer. By
               | replicating the data, you can pull it from multiple
               | replicas at once (I.e. there are more seeders). This also
               | protects against someone suddenly getting throttled, or
               | just having a power outage. The protocol should
               | prioritize donors with high bandwidth over vast storage
               | space.
               | 
               | Feel free to DM me on Twitter if you'd like to toss
               | around some design ideas more seriously, and thank you
               | for trying to build this.
        
               | stavros wrote:
               | Thanks, this is exactly what I wanted! I'll DM you on
               | Twitter now!
               | 
               | Edit: Looks like Twitter wants me to pay to DM, I'll
               | email you instead.
        
               | sillysaurusx wrote:
               | Wait what? My DMs are open. But people do occasionally
               | say they have trouble DMing me.
        
               | stavros wrote:
               | Thanks! It said I need to be verified to DM people who
               | aren't following me.
        
               | sillysaurusx wrote:
               | Thank you, that's helpful to know. And frustrating. I see
               | why Twitter did that, because bots, but I was willing to
               | wade through the crap to find the gems. Which do get
               | sent.
               | 
               | I responded with some telegram info if that helps.
        
               | the8472 wrote:
               | > The protocol should prioritize donors with high
               | bandwidth over vast storage space.
               | 
               | If you're doing this in bittorrent then you might want a
               | client that's configured to optimize for a different goal
               | than most torrent clients.
               | 
               | Potential goals, somewhat conflicting:
               | 
               | A) keep data with a low mirroring degree available.
               | either this needs to be centrally coordinated or some
               | sort of randomized algorithm where clients pick
               | underseeded torrents but not everyone picks the same
               | 
               | B) bandwidth matching. to not consume more resources than
               | are provided a client maybe should only download 1 piece
               | of data for every N times it uploaded any piece. This is
               | much less greedy than what you'd have in a normal torrent
               | client but ensures that caches themselves don't take up
               | much bandwidth compared to users who actually want to
               | download something. Otherwise a misconfigured cache (e.g.
               | behind NAT) could accidentally always download data
               | without ever giving much back.
        
               | vasco wrote:
               | Haven't you described IPFS minus crypto shenanigans?
        
               | stavros wrote:
               | Not exactly, IPFS doesn't tell you what to download (you
               | select what to download) and thus can't push the rarest
               | material to you. There are many similarities, but this is
               | much more suited to making large datasets more
               | resilient/accessible.
        
         | golol wrote:
         | Any entity should be allowed to read, process and update its
         | internal state using published media. Only your productions
         | should be judged based on copyright law.
        
         | pixl97 wrote:
         | The biggest theft in history was you being taught the alphabet,
         | numbers, and words without having to pay a creator for them.
         | This mistake will be reversed. All knowledge will be licenced,
         | all words will be owned.
         | 
         | Short summary of The Right to Read, and evidently the world you
         | want.
        
       | bomewish wrote:
       | Oh my gosh. Duxiu is an absolutely incredible resource. The
       | prospect that it will be fully searchable in a year is
       | incredible!!
        
         | pilimi_anna wrote:
         | It really is
        
       | bberrry wrote:
       | Do LLMs really encode knowledge regardless of language, and tap
       | that knowledge irrespective of which language you query in?
       | 
       | My mental model of LLMs being next-word predictors with a long
       | context window suggests they don't. Are there any papers on this?
        
         | gabipurcaru wrote:
         | this is fairly easy to test for yourself. What did you try?
        
           | bberrry wrote:
           | I don't think it's easy at all without access to the training
           | data. I could ask about some information I find obscure in my
           | native language but I can't be confident someone didn't write
           | about it in English on e.g. reddit.
        
           | sebzim4500 wrote:
           | How exactly are you proposing he tests this without access to
           | hundreds of thousands of dollars worth of compute? Toy models
           | don't work for this kind of thing, small language models
           | behave qualitatively differently from large ones.
        
         | stevenhuang wrote:
         | you may disabuse yourself this mental model by learning about
         | the concept of universal function approximators and the
         | possibility that intelligence itself can be approximated.
        
         | golol wrote:
         | SGD in high dimensions is good at finding efficient,
         | generalizing models. Naturally, it is more efficient to learn a
         | representation of the abstract concept of a table rather than
         | learn representations for the hundreds of different words for
         | tables. I don't have reference but for me it feels very
         | inthuitive that the former is the case.
        
         | DougBTX wrote:
         | A big area of research in NLP is translation between natural
         | languages, I'd go so far as to say that "translation tool" is a
         | better mental model than "token predictor". For example, think
         | of summarisation as translation between verbose language and
         | concise language, or code generation as translation between
         | English and Python.
        
           | HPsquared wrote:
           | They are called "transformers", after all.
        
         | LudwigNagasena wrote:
         | What is a better way of predicting the next word if not
         | knowledge encoding?
        
         | the8472 wrote:
         | Predicting the next word is hard. To do this more efficiently
         | than a dumb lookup table (the chinese room thought experiment)
         | or a markov chain you basically need to build a world model.
         | This model sits in the deeper layers of the LLMs (the deep in
         | deep learning). Turning it into english or chinese only happens
         | closer to the output layers.
         | 
         | https://gwern.net/scaling-hypothesis#why-does-pretraining-wo...
        
         | sebzim4500 wrote:
         | Anthropic has a recent paper that talks about this [1]. See
         | figure 16 in particular. The larger the model, the more it
         | generalizes between languages.
         | 
         | On the other hand, no matter how large the model they struggle
         | to generalize between sentences of the form "George Washington
         | was the first US president" and "The first US president was
         | George Washington".
         | 
         | Clearly generalization behavior is unintuitive, you can come up
         | with a post hoc explanation (weights are independent between
         | early and late layers) but I doubt anyone would have thought
         | ahead of time that models would have an easier time
         | generalizing between languages than between slightly different
         | ways of saying the same thing in English.
         | 
         | [1] https://arxiv.org/pdf/2308.03296.pdf
        
         | kolinko wrote:
         | There was a nice paper recently showing that even people encode
         | the knowledge differently depending on which language they use.
         | 
         | According to that paper, people made decisions more emotionally
         | in their native languages compared to their second language -
         | where they tend to be more logical.
         | 
         | In case of LLMs there were some nice cases where GPT gave
         | different replies depending on the language of the query.
         | Factual information was roughly the same, but in some cases the
         | model gave totally different replies depending on the language
         | of the query.
         | 
         | As a bilingual speaker - my experience is that the replies are
         | very similar regardless of whether I speak in Polish or
         | English, or mix them both within the same sentence.
        
         | justanotherjoe wrote:
         | you say it like predicting next word is an easy task.
         | Calculating the permutations with statistics alone would
         | probably require more compute than the entire universe to the
         | power of the entire universe. And there's just not enough data,
         | even all the internet wont be enough. We abandoned these
         | methods long ago.
         | 
         | Yes, LLMs really encode knowledge. This is not up for debate,
         | and it's obvious to anyone working in AI. It outputs words and
         | input words, yes, but the magic is what is in between. The
         | words are broken down immediately at the first layer. Next it
         | involves 100s of billions of parameters doing god knows what.
         | But it's safe to assume they encode the base reality as best it
         | can. It is what i call 'language-induced reality model'.
         | 
         | The task of next word predictor in a humongous dataset turns
         | out, is best solved by creating an inner representation of the
         | real world. It's still not a perfect model since it's just
         | induced by language though, so obviously it has limitations.
         | But it's really modeling it, and with higher fidelity than
         | people imagined you can learn the world through words alone.
        
         | PeterisP wrote:
         | Standard large transformers trained on corpora of multiple
         | languages will generally perform next-word prediction in
         | language A by using information that was only seen in training
         | data in language B, demonstrating that they have managed to
         | implicitly learn a capability for translation and/or
         | multilingual perception.
        
       | JCharante wrote:
       | Chinese languages have a higher information density which makes
       | me hypothesize that tokenizing works better than in Romance
       | languages which can lead to more efficient LLMs.
        
         | teruakohatu wrote:
         | Right now English language LLMs favour more tokens over fewer
         | tokens and more unique tokens. For example in GPT4, "
         | Altogether now" is four tokens.
         | 
         | I don't think visual information density is much help.
        
           | JCharante wrote:
           | I'm just an observer to the space, but I imagine each chinese
           | character (or each word) being a token would be pretty
           | effective, especially as inputting/outputting more tokens
           | requires more vram. Additionally Chinese doesn't have
           | conjugations which I imagine makes tokenization easier.
           | 
           | I think it's easier to connect semantic meaning between these
           | tokens/charcaters, because you can relate a character to
           | other characters by their presence in words (most words are 2
           | characters long (compound words), although there are
           | sometimes words like AB and BC which mean the same thing, but
           | in conversations people will just say A or C and the meaning
           | (AB or BC) is inferred by the context of the conversation, so
           | in a semantic meaning database you would want a connection
           | between A and AB, and to BC & C). You could also have a
           | dimension for each radical that isn't a phonetic component.
           | Radicals do link vaguely related words together, and I think
           | this is a powerful language feature missing in English that
           | could be exploited in the NLP field.
           | 
           | I'm not a fluent or proficient speaker, so maybe I'm not the
           | best person to explain things, but before OpenAI's Whisper,
           | speech-to-text in WeChat in Mandarin was way better than any
           | English speech-to-text I've ever used. The structure of the
           | Chinese language and its large userbase would make me bet
           | long on Chinese NLP progress over the west.
        
             | vasco wrote:
             | Imagine losing WW3 because your language is less efficient
             | to store and query than the enemy, so your war-llms need
             | more resources and are less effective.
        
               | stavros wrote:
               | My understanding is that all spoken languages are roughly
               | equally efficient at data transmission per unit of time.
               | I'm not sure how long it takes someone to read a text in
               | Chinese over a text in English, though.
        
               | JCharante wrote:
               | Per unit of time is true, because languages with a lesser
               | information density will be spoken at a faster rate in
               | order to compensate (see: Spanish speakers speaking way
               | faster than English speakers). But LLMs don't ingest data
               | per unit of time, they ingest data via text.
        
               | stavros wrote:
               | Sure, but, depending on your representation, you can only
               | have so many different tokens, and if you use, say,
               | pinyin, then you don't have much of an advantage over
               | English.
               | 
               | It'll be interesting to see how LLMs do when trained on
               | concepts (ie the Chinese alphabet), rather than sounds,
               | though.
        
               | JCharante wrote:
               | > Sure, but, depending on your representation, you can
               | only have so many different tokens
               | 
               | From my understanding (which only comes from reading HN),
               | unique token count isn't an issue that LLMs run into. If
               | it is, then that would be a bummer for the possibilities
               | of exploiting all the features of the Chinese language.
               | 
               | > and if you use, say, pinyin
               | 
               | Well yeah pinyin would be inefficient because you're
               | stripping away all the "built in" semantics of a Chinese
               | character to only focus on the pronounciation.
        
               | stavros wrote:
               | I imagine you can always expand your token count, but
               | then is that very different from syllabic alphabets (that
               | encode multiple syllables in one character)?
               | 
               | Granted, that doesn't apply to Chinese, which encoders
               | concepts, so that's interesting to see in LLMs.
        
               | JCharante wrote:
               | I think so, because even if you're encoding multiple
               | syllables, that's just the pronounciation.
               | 
               | I've learned languages with alphabets and glyph scripts
               | like Chinese, and *what I don't like about alphabet
               | languages like English is that the letters themselves
               | provide little context to their meaning, although you
               | could learn the latin root words and guess from there. Of
               | course you know how to say it, but you don't know what it
               | means. With Chinese, you might guess what it means, but
               | you don't know how to say it.* The characters have more
               | meaning in them, although this isn't the case for every
               | word, which is something non-speakers assume.
               | 
               | In short, from my experience learning Chinese was way
               | easier than learning Vietnamese, which have historical
               | ties together since Vietnamese used a fork of Chinese
               | script until the early 1900s when they transitioned to
               | latin script in order to improve literacy rates. Sure, it
               | improved literacy rates because to read you only need to
               | know how to pronounce, but it doesn't mean glyph
               | languages are harder to learn the meaning/semantics.
        
             | czl_my wrote:
             | Interestingly BaiChuan2 didn't just tokenize individual
             | characters
             | 
             | https://huggingface.co/baichuan-inc/Baichuan2-7B-Chat
        
             | visarga wrote:
             | Text tokens by their construction are more like the Chinese
             | ideographic characters, so no, there would not be a big
             | difference between tokenized English and Chinese. The
             | number of unique tokens is decided upfront - something like
             | 50K - so, much more than basic vocabulary. English tokens
             | can even be 2 word phrases.
             | 
             | Look up Byte-Pair Encoding (BPE) for a fascinating and
             | simple algorithm that explains the process of choosing the
             | tokens. It shows why it doesn't matter how token dense the
             | source language is.
        
             | kolinko wrote:
             | Vram is dependent mostly on the parameter count, not on the
             | number of input/output tokens, if I remember correctly.
             | 
             | Also, in case of english language, it's one token per one
             | word (more or less), so it's the same as in Chinese -
             | assuming both LLM tokenisers were geared towards their
             | native language.
             | 
             | The only issue is when you have an tokeniser geared towards
             | western languages, and you try to to use the same tokeniser
             | on a different group of languages - then a single word in a
             | language foreign to the tokeniser would have multiple
             | tokens.
             | 
             | But that has nothing to do with the underlying structure of
             | the language.
             | 
             | In other words - you wouldn't really see a difference
             | between an input in chinese compared to english after the
             | text gets tokenised. It's rougly the same amount of tokens,
             | and the underlying parameter count would be also similar.
        
         | jabbany wrote:
         | Not particularly? The number of "words"/distinct concepts is
         | not that much different between languages. Having more
         | characters just makes the words look shorter...
         | 
         | That said, Chinese is an isolating language, so there is little
         | information within a word to hint at the rest of the sentence
         | (no plurality, tense etc.). Which could be better or worse for
         | language models. Or it might not matter in the end given modern
         | tokenization also uses word "fragments" for Western languages.
        
         | kolinko wrote:
         | I wouldn't be sure of that:
         | 
         | - if by efficiency you mean the speed of operation, then it's
         | the network size that determines it
         | 
         | - the input token / vector size is not dependent on the
         | language
        
       | Havoc wrote:
       | I'm guessing one of the big Chinese firms will outbid everyone.
       | They're obviously trying to catch up with OpenAI and throwing
       | whatever money necessary at this makes sense as a result
        
         | sillysaurusx wrote:
         | That's ok. It's the expected result in a market economy.
         | 
         | America is throwing out their chances of making open source
         | LLMs. Copyright holders are demanding license fees for
         | training. That's analogous to someone demanding licensing fees
         | _before_ you make a YouTube video. No one would be able to do
         | it for free. Whereas it was completely possible to train a high
         | grade LLM for free (clusters are surprisingly accessible for
         | researchers) and it wasn't till recently that you had to worry
         | about being sued for it.
         | 
         | Net result: open source LLMs die, except for companies that can
         | open source their smaller (lamer) models as an upsell for the
         | real ones that anybody cares about. That's not a world where
         | open source makes a big impact. That's a world where
         | (metaphorically) GPL software is subservient to business
         | interests for the rest of eternity. Say what you will about
         | whether that's true, but no business has influence over Emacs,
         | and it's fantastic, powerful software. No one will be able to
         | make the equivalent open source fantastic LLM in America at
         | this rate.
        
           | visarga wrote:
           | Doesn't have to be made in America, you can download a LLaMA
           | or Mistral in minutes anywhere. A Chinese base model can be
           | fine-tuned anywhere.
        
           | colinsane wrote:
           | i'm not sure i follow. you're saying that no company (except
           | the small ones) will be able to open source a LLM. but then
           | you cite software created _outside of the traditional company
           | structure_ as an example of what we 'll never have in the LLM
           | space. doesn't your example negate the premise?
        
             | sillysaurusx wrote:
             | Not at all. Maybe it's lost to time, but most of the
             | important models were created by academics, not companies,
             | up till recently. GPT-J for example was trained by one
             | person acting alone (Ben Wang).
             | 
             | I fine tuned GPT 1.5B on chess games. AI Dungeon fine tuned
             | on fantasy novels. All of this type of work will become
             | impossible with the specter of lawsuits hovering over their
             | heads.
             | 
             | EDIT: also most impactful older models were by one person
             | (e.g. Yolo). What I like about ML is that lone wolves can
             | have a big impact.
        
           | carlosjobim wrote:
           | What would hinder an open source pirate LLM?
        
             | sillysaurusx wrote:
             | A bunch of factors. One is that you'd have to keep your
             | identity private, but credibility is how you get access to
             | resources. And resources are necessary to train anything.
             | 
             | Take TRC for example. They give people access to TPUs in
             | exchange for being cited. But if they were cited as
             | facilitating large scale piracy, they probably wouldn't be
             | happy. It could even lead to a lawsuit on the grounds of
             | facilitating copyright infringement, which will likely be
             | the charge against me if someone gets mad enough to sue me
             | directly. (I never distributed anything, but that doesn't
             | matter if they can prove facilitation.) And TRC is an even
             | juicier target for lawsuits since Google is a giant loot
             | box of money for them.
        
             | pixl97 wrote:
             | The coordination problems faced by all illegal entities.
             | 
             | Piracy generally means taking a full fledged product of
             | someone elses and using it. Seemingly in your description,
             | you're taking the data illegally and then doing all the
             | compute on it. The amount of compute needed in this case is
             | staggering, so you are back to coordinating with other
             | people. Any one of those individuals turning against the
             | group would likely doom the group, hence it's a high risk
             | operation.
        
         | mensetmanusman wrote:
         | Chinese firms can't afford to hire the number of censors
         | required to satisfy the very proud leaders of the CCP.
        
       | codedokode wrote:
       | AI is a technology that is going to bring ridiculous profits,
       | billions of dollars. Why don't AI researchers want to pay for the
       | books and for the work to write them? They want to train their
       | models for free, and then sell access for money like OpenAI.
        
         | sillysaurusx wrote:
         | Imagine saying that someone should pay for the right to make
         | open source software, before you've made anything, or proven
         | any kind of market value. In a world with power law
         | distributions, the vast majority of people will not come close
         | to even millions in profit. At one point I was hoping just to
         | run my own GPT API to self fund my measly $7k/mo burn rate.
         | Then I could live life happy as a clam doing research all day
         | in my little shell.
         | 
         | Such a dream is completely impossible when rightsholders demand
         | everyone pay for training data, rather than sharing in the
         | profits of the result. It's backwards. The actual market profit
         | is what matters. But authors are frothing at the mouth, upset
         | that their work is being used at all without their explicit
         | consent. And some of those authors, like Sarah Silverman and
         | more recently former governor Mike Huckabee, are going after
         | LLM companies and demanding injunctions preventing them from
         | using books at all, under any circumstance, unless they go
         | through authors. Good luck going through 100,000 different
         | authors.
         | 
         | The net result is that we're headed for a world where the
         | YouTube of LLMs will dominate the space, and no one else will
         | be able to run their own platforms. I've done everything
         | humanly possible to try to steer us away from that outcome --
         | you like using stable diffusion yourself? Imagine if no one
         | released stable diffusion in the first place. But things are
         | not looking good.
         | 
         | Lawsuit against Eleuther for books3, courtesy of Huckabee:
         | https://news.ycombinator.com/item?id=37962244
         | 
         | Example of authors being furious, though thankfully with an
         | exception for academic curiosity:
         | https://news.ycombinator.com/item?id=37949012
         | 
         | Twitter is completely merciless. Digging up some examples to
         | illustrate the point...
         | 
         | Here's one: look through the replies of someone who released a
         | 3D dataset.
         | https://x.com/mattdeitke/status/1678855859089326080?s=61&t=j...
         | "Eat shit Matt" (5O likes) "Hope you enjoy the storm" "you make
         | me sick" and on and on. I can tell you firsthand that this
         | isn't unusual; it's the default reaction now.
        
           | stavros wrote:
           | The good thing about LLMs (at least when compared to YouTube)
           | is that "all" you need to train an LLM is processing power,
           | so it's easier now (and will be much easier in the future) to
           | do it illegally. Much easier than running a YouTube
           | competitor illegally for any length of time.
        
           | vasco wrote:
           | > Such a dream is completely impossible when rightsholders
           | demand everyone pay for training data, rather than sharing in
           | the profits of the result. It's backwards. The actual market
           | profit is what matters
           | 
           | I like the way you think, can I also get free books at the
           | store if I'm just reading them for learning and will only
           | make money with the knowledge later, or if I promise that I
           | won't put it in practice?
           | 
           | I did a similar deal in university but had to return books to
           | the library every week so I like your plan better.
        
             | pwagland wrote:
             | One big difference with your university analogy is that the
             | university _did_ pay for the resource, often way more than
             | you would as a private citizen, since it is for sharing
             | purposes.
             | 
             | Even worse, for e-books, libraries are often limited to how
             | often they can lend the book out, and this was done since
             | physical books "degrade" over time, and so have a limited
             | lend-life, the idea is that the same should hold true for
             | digital books as well...
             | 
             | I get the point behind your sarcasm of the first sentence,
             | since that is the essence of the problem.
        
             | TeMPOraL wrote:
             | Well yeah, people sometimes read whole books in a store.
             | Though this use case is done better in facilities that
             | specifically support it - libraries.
             | 
             | If you really need to take the books back with you, that's
             | what the library copy machine is for.
        
           | sensanaty wrote:
           | Why do you deserve to reap the benefits of their labour while
           | not giving them a cent? Just saying "but open source!" Isn't
           | a valid argument, everyone who wants access to someone else's
           | work should be forced to pay, this includes the current big
           | players that have so far been circumventing paying people for
           | their work.
           | 
           | It's attitudes like these from the AI sycophants that makes
           | people despise the technology and anyone involved in any of
           | it. I don't give a shit how difficult not having free
           | unfettered access to everyone else's work makes it for you. I
           | don't care that we're supposedly holding the world back (from
           | infinite machine-generated spam made for the sole purpose of
           | making everyone else's lives maximally miserable).
        
             | sillysaurusx wrote:
             | In general, the goal of labor is to enrich the society you
             | live in. Whether it's capitalism or socialism, the theory
             | is the same: the richer your society, the happier you'll
             | be, even if you're poor. Objectively, the poorest in
             | America live far better than the poorest in poor countries.
             | 
             | Then there is the harm argument. What is the harm to
             | authors of letting AI models do this? Models aren't
             | replacing authors any time soon. And even if they were, you
             | won't be able to stop it; precisely zero times in history
             | have people stood against a wave of new technology.
             | Technology wins 10 out of 10 times in the long run.
             | 
             | I am skeptical that authors deserve anything for merely
             | having a transformative use of their work. But I'm also
             | sympathetic. The key is for authors to get a slice _of the
             | actual market value_. Right now authors are shutting
             | everyone down merely for using their work at all, with no
             | consideration of whether anyone is even profiting.
             | 
             | Yet even when authors deserve a cent, how do you decide
             | which authors get how many cents? Do all profits go to
             | copyright holders, as with YouTube? Then there's no room to
             | run a service. Is half fair? Before or after company
             | expenses? And whatever is allocated, is it distributed
             | evenly among all 100,000 authors, whether they're Stephen
             | King or just an enthusiastic contributor to AO3? $1M in
             | profits distributed to 100,000 authors is a whopping $10
             | per author. That'll buy them a Happy Meal and fries. It's
             | far more lucrative to make up imaginary damages and sue
             | them in court.
             | 
             | It's immensely complicated, and it's not at all clear that
             | it's evil or even exploitative to train AI models on other
             | people's work.
        
         | visarga wrote:
         | > Why don't AI researchers want to pay for the books and for
         | the work to write them?
         | 
         | You get paid in exchange - you give them your text, they give
         | you their AI. Have you tried Mistral-7B, it can do amazing
         | feats on a 5 year old GPU. Didn't imagine my old GPUs were
         | going to get so smart someday.
         | 
         | LLMs by their nature give every user what they need, adapting
         | and customizing the repository of human knowledge to a
         | particular situation. That's how they pay back, they adapt to
         | serve our needs.
         | 
         | Now the authors can use AI to help themselves - brainstorm
         | ideas, critique, first drafts, etc. But in order to benefit
         | they got to use it first.
        
         | userbinator wrote:
         | _They want to train their models for free, and then sell access
         | for money like OpenAI._
         | 
         | Counterpoint: Someone who teaches himself how to do something
         | entirely via freely borrowed books from the library, and then
         | makes a living out of those learned skills, is normally not
         | considered stealing.
        
           | stevewallis wrote:
           | But... how do you think the book got into the library? The
           | state buys the book from the author's publisher. The author
           | gets paid their royalty. This happens with your tax dollars.
           | This is a good thing, the library system allows authors to
           | receive payment to provide a public good, that can educate
           | citizens who can then make a living. Stealing their work to
           | train an LLM to make a living is not the same thing.
        
             | PeterisP wrote:
             | Okay, then let's have libraries train LLMs on the whole
             | collection of the books that they have legitimately
             | obtained. I'd assume that for this particular collection of
             | Chinese printed works, the National Library of China would
             | also own a copy of every one of them, so the end result
             | would be literally be the same.
        
           | sensanaty wrote:
           | And a computer isn't a human luckily, so this argument
           | doesn't work here.
        
         | raincole wrote:
         | People will write complex moral and legal reasoning.
         | 
         | I'll tell you the simple answer. They don't want to pay for the
         | books because they can get it for free. Simple as that.
        
       | Alifatisk wrote:
       | I wonder if LLMs will enable us to access information in chinese
       | much easier now by acting as an interface.
        
       | popcalc wrote:
       | The guy running this site is undoubtedly intelligent but also
       | foolhardy enough to where it's going to be hard to feel sorry
       | when his door gets broken down. He started this project after
       | seeing how dangerous of a situation zlib's spotlight put them in,
       | yet he tried to replicate their success. He succeeded but I
       | recommend he doesn't take being a free man for granted.
        
         | sillysaurusx wrote:
         | Others will carry the torch. Silk Road was proof of that. Of
         | course, it's little consolation to Ulbricht, but he deserved to
         | be imprisoned for ordering hits. (Incidentally, I was skeptical
         | of that claim till
         | https://youtu.be/GpMP6Nh3FvU?si=q0KXzP2FNJDin2tW.)
         | 
         | The benefit of Anna's Archive is that they have goodwill on
         | their side. They are trying to contribute to academic research
         | directly, whereas libgen seems to prioritize profit (and has
         | to, to survive).
         | 
         | Anna's Archive does that too. The "donations" are actually
         | paying to access features. But by proactively reaching out to
         | the academic community like this, they'll make a lot of
         | powerful allies.
         | 
         | EDIT: on second thought, how does libgen monetize? I thought
         | they ran ads, but I just realized I didn't verify that.
         | 
         | Whoever runs a library like this for altruistic reasons stands
         | to win. The desire to get rich is most of the problem with
         | doing shady things scalably. One can argue that it's the
         | risk/reward tradeoff, but I suspect as technology makes
         | everything more accessible, an operation like this only costs a
         | few thousand a month. That's achievable while flying under the
         | radar long enough to hopefully avoid getting in a lot of
         | trouble.
         | 
         | Of course, as with death note, everything comes down to: don't
         | get caught.
        
           | OfSanguineFire wrote:
           | > how does libgen monetize
           | 
           | The original LibGen, under anonymous and altruistic founder
           | bookwarrior, did not aim to monetize but was built purely on
           | volunteer effort. Some of the LibGen domains, however, are
           | forks that put up advertisements.
        
             | sillysaurusx wrote:
             | That's so cool.
        
           | qingcharles wrote:
           | It's hard to know if they are running it for altruistic
           | reasons or money or both. Probably both. You _can_ make a lot
           | of money. I ran a torrent site in the 00s and it made  >$10m
           | in donations.
        
         | squigz wrote:
         | Interesting choice of pronouns to refer to the operator of
         | _Anna 's_ Archive
        
           | GaggiX wrote:
           | It seems naive to me to think that Anna is one person and not
           | a team of people.
        
             | simonw wrote:
             | So "The guy running this site" should be "The people
             | running this site" then.
        
               | GaggiX wrote:
               | Yeah exactly, they were both being naive about it.
        
       | RecycledEle wrote:
       | The Chinese system where all books are in a central database
       | gives them huge advantages in training AI.
       | 
       | I just realized that western companies are screwed because of the
       | people who wanted to make a buck by creacting artificial
       | scarcity.
       | 
       | I have been predicting for decades that IP laws would bring about
       | the end of our civilization. I did not know how it would happen.
       | Now I know.
        
         | dragonelite wrote:
         | That way to dooming the west will be fine the rest of the world
         | will just do better then they have been doing the last 3
         | centuries. It will create a level playing field and the current
         | tier like system.
         | 
         | We are going back to the time where GDP was just how many
         | farmers did your country have. But in 2000s it will be how many
         | STEM people does your country have and how much IP one can
         | create and license. tech IP is like a global rent from
         | everybody that consumes tech.
        
       | mensetmanusman wrote:
       | The Chinese have a very different cultural response to copying,
       | and often see it as a form of honoring.
       | 
       | They would be more than happy if everyone on earth had every
       | Chinese work in existence.
        
         | faitswulff wrote:
         | That seems essentialist. Chinese incentives are aligned with a
         | lack of IP protection right now. For instance, IIRC, bunny's
         | explanation of gongkai
         | (https://www.bunniestudios.com/blog/?p=4297) explains that
         | Chinese companies freely trade chip designs because they make
         | money off of manufacturing, regardless of who designed it.
        
           | NotSuspicious wrote:
           | This seems to me a much more moral way of doing things.
           | Actually making the stuff is what's important.
        
             | error_logic wrote:
             | Patents and copyright are intended to encourage publishing
             | of information (and profitability of individuals so they
             | can establish themselves with economic mobility and unlock
             | more potential investments for the world) rather than
             | keeping it as trade secret. That doesn't mean they always
             | do their job, but neither will there always be people
             | running a company or a nation who can maintain such sharing
             | in the face of competitive pressures during different
             | times.
        
           | yorwba wrote:
           | > Chinese incentives are aligned with a lack of IP protection
           | right now.
           | 
           | Even that is essentialist. In some areas, sharing IP is
           | beneficial to the originators and incentives are aligned. In
           | other areas, not so much.
           | 
           | Duxiu makes money selling access to their collection of
           | scanned books. Their incentives are not aligned with having
           | others use those scans for free. Currently, such cases seem
           | to be mostly handled under the most general provisions of the
           | Anti-Unfair Competition Law (Fan Bu Zheng Dang Jing Zheng Fa
           | ), but new amendments are likely to make it more explicitly
           | illegal https://www.whitecase.com/insight-alert/china-
           | releases-draft... ("Improperly obtaining or using another
           | business operator's commercial data")
        
         | livinglist wrote:
         | Oh no, there's a huge difference between Shan Zhai (copy) and
         | Zhi Jing (pay tribute)...
        
       | richwater wrote:
       | We shouldn't help the chinese with anything. They steal from us
       | daily.
        
       | xeckr wrote:
       | Does "Chinese nonfiction" include Mao's bibliography and "Xi
       | Jinping thought"?
       | 
       | Do you realize how all-consuming China's censorship apparatus is,
       | and that this censorship introduces all sorts of biases in their
       | literary corpus which you now want to be used as training data
       | for LLMs?
        
         | staticman2 wrote:
         | That's why you fine tune models, right? To make sure the
         | undesirable things in their data set don't lead to undesirable
         | output.
        
           | xeckr wrote:
           | Fine-tuning will never get rid of all biases introduced by
           | the training data, some of which are subtle enough that you
           | won't even know that you should apply fine-tuning techniques
           | to try to mitigate them.
           | 
           | For example, did you know that there is a trend for _physics
           | textbooks_ to be dedicated to Xi Jinping? If you didn 't know
           | this, because who has time to sift through millions of books
           | to find undesirable biases, isn't it clear that your LLM will
           | develop subtle but favourable biases towards the CCP?
        
       | tway223 wrote:
       | This collection has been on the internet for quite a while,
       | likely started around 2015-ish. It is highly duplicated and I
       | suspect the total number is around 4 million books. Still a lot.
       | 
       | The source was from a company named DuXiu, or previously
       | SuperStar. They collaborated with the libraries around China and
       | scanned their collections since early 2000-ish. Before that I
       | think they just bought some junk books from recycling stations
       | based on the quality of early samples.
       | 
       | Many of the books are translated versions of the textbooks from
       | the west (most likely the US) and many are pure political
       | propaganda junk. Some literature and history stuff which were
       | published when censorship wasn't so extreme.
       | 
       | Many of the Chinese tech companies should have access to this
       | collection (especially Baidu for sure) but the books were not
       | censored based on today's standards so I doubt any of them would
       | openly use them not only due to the copyright issue but also the
       | political risks.
        
       ___________________________________________________________________
       (page generated 2023-11-05 23:01 UTC)