5 ms·

So what I got from this is that GPT was trained on a dataset that biased in English contents. Is that right? I think even human has to spend extra energy to sp

So what I got from this is that GPT was trained on a dataset that biased in English contents. Is that right? I think even human has to spend extra energy to speak a language they were not born with, no matter how fluent they are in this language. I don't know about natural multilinguals.

kevingadd 4y ago
Should the cost really be 15x? Or even 5x? In this case, it's not even a question of whether the network is better at English, it's that the cost to communicate with it at all in other languages is higher. Once you pay that cost you now have to deal with the network potentially generating lower quality results for prompts in non-English languages too, which raises the actual cost of doing something with GPT beyond 15x since you probably will need more attempts.
- viscanti 4y ago
  Because there's so much more English language for them to train on relative to most other languages, they're able to do some optimizations for English that they can't elsewhere. Should they not be able to implement optimizations for cases where they have the data volume to do so?
  - kevingadd 4y ago
    "there's so much more English language for them to train on relative to most other languages" is an interesting assertion. There are billions of people on earth speaking languages other than English and they have access to the internet. Are you sure it's not just the case that we didn't scrape that data? Everyone has to choose what data to train on, you can't train against The Entire Internet, it's a limitless amount of data. But it becomes an intentional choice with consequences, like the 15.77x seen here.
    - JCharante 4y ago
      but training against the entire Internet would still be biased towards English because English is the dominant language used on the Internet.
    - MichaelZuo 4y ago
      > Everyone has to choose what data to train on, you can't train against The Entire Internet, it's a limitless amount of data. Isn't that exactly how OpenAI managed to 10x GPT 3.5 with GPT 4.0?
  - famouswaffles 4y ago
    Both of you are kind of misunderstanding a few things. Data used to train the tokenizer is entirely separate from data training the LLM. The tokenizer used to train GPT-3 was old, inefficient and targeted at tokenizing English. That's pretty much all there is to it. It's possible to train a tokenizer that is more efficient and more including of other languages. GPT-4's tokenizer is already far more efficient though still weighted to English.
    - viscanti 4y ago
      > GPT-4's tokenizer is already far more efficient though still weighted to English. Right. It's a general question. Should they be allowed to take the kinds of optimizations they can with tokenization when it's a function of how much data they can use, even if that means some languages get more optimization than others? Or should users of those languages that could be optimized effectively pay a tax out of some sense of fairness?
- user_named 4y ago
  What makes you think there's a "should"?
  - cool_dude85 4y ago
    There's always a should. Society gets a say in what people and corporations can and can't do in (at the very least) the form of laws. There's your should right there.
    - pixl97 4y ago
      Which society? I mean OpanAI is a US company is unsurprisingly going to mostly communicate in English. Are we counting all societies, if so should our software cater to all their demands, language and/or culturally demanded?
- famouswaffles 4y ago
  Both of you are kind of misunderstanding a few things. Data used to train the tokenizer is entirely separate from data training the LLM. The tokenizer used to train GPT-3 was old, inefficient and targeted at tokenizing English. That's pretty much all there is to it. It's possible to train a tokenizer that is more efficient and more including of other languages. GPT-4's tokenizer is already far more efficient though still weighted to English.
terafo 4y ago
Nope, it's not about dataset. It's just bad tokenizer. Korean has couple of dozen of symbols in it's alphabet. Cyrillic languages have less than 50 symbols in total. Hiragana is 46 symbols. GPT-4 has 32k tokens IIRC. Including most significant alphabets would take less than a thousand.
- moelf 4y ago
  but the tokenizer is dataset-driven... it tokenize the most common pattern in your dataset to improve efficiency, so it's 100% about dataset?
  - sillysaurusx 4y ago
    There's dataset during training, and dataset for the tokenizer. The confusion here is that people are talking about the former, but you're correct that it's the latter. Remember, OpenAI's tokenizer was created in an era when 125MB was considered large for a language model. It's hard to fault them for making something that lasted four or five years.
    - dragonwriter 4y ago
      > Remember, OpenAI’s tokenizer was created in an era when 125MB was considered large for a language model. GPT-2 and GPT-3 have different vocabularies and maximum token #s, which (even if the tokenizer architecture is the same) implies a different tokenizer model. GPT-3.5 might share the GPT-3 tokenizer, but even then I’d expect GPT-4 to have its own. But even if they are using the tokenizer from GPT-3, its not from “an era when 125MB was considered large for a language model”.
      - sillysaurusx 4y ago
        Actually, GPT-3's tokenizer is the same as GPT-2. https://datascience.stackexchange.com/a/109483 https://datascience.stackexchange.com/a/109483 You had me questioning myself for a minute. (The vocab size is still 50257. Even rounded up to a multiple of 128 for better sharding across the vocab embedding, only the first 50257 are used.) Believe it or not, 125M was large at the start of the GPT-2 era. No one knew LLMs could do anything interesting, let alone that they'd change the world.
tlrobinson 4y ago
I think yes, but more precisely the tokens were chosen to optimize training on a dataset that's biased to English content. I am curious how the token set affects quality of responses, ignoring the factors related to token count mentioned in the post (cost, prompt expressivity, latency, etc) Is it always better for the token set to be "native" to the majority of the training dataset and prompts/completions, or is it possible there's some "intermediate representation" (in compiler terms) that would be better?
- k8si 4y ago
  I don't know what you mean by compiler terms but basically, worse tokenizer = worse LM performance. This is because worse tokenizer means more tokens per sentence so it takes more FLOPs to train on each sentence, on average. So given a fixed training budget, English essentially gets more "learning per token" than other languages.
GaggiX 4y ago
The article only clarifies that the dataset used to train the tokenizer is baised, not the entire dataset used by the GPT model.
famouswaffles 4y ago
Data used to train the tokenizer is entirely separate from data training the LLM. The tokenizer used to train GPT-3 was old, inefficient and targeted at tokenizing English. That's pretty much all there is to it. It's possible to train a tokenizer that is more efficient and more including of other languages. GPT-4's tokenizer is already far more efficient though still weighted to English. You can test it here https://tiktokenizer.vercel.app/ https://tiktokenizer.vercel.app/

kevingadd 4y ago

Should the cost really be 15x? Or even 5x? In this case, it's not even a question of whether the network is better at English, it's that the cost to communicate with it at all in other languages is higher. Once you pay that cost you now have to deal with the network potentially generating lower quality results for prompts in non-English languages too, which raises the actual cost of doing something with GPT beyond 15x since you probably will need more attempts.

viscanti 4y ago
Because there's so much more English language for them to train on relative to most other languages, they're able to do some optimizations for English that they can't elsewhere. Should they not be able to implement optimizations for cases where they have the data volume to do so?
- kevingadd 4y ago
  "there's so much more English language for them to train on relative to most other languages" is an interesting assertion. There are billions of people on earth speaking languages other than English and they have access to the internet. Are you sure it's not just the case that we didn't scrape that data? Everyone has to choose what data to train on, you can't train against The Entire Internet, it's a limitless amount of data. But it becomes an intentional choice with consequences, like the 15.77x seen here.
  - JCharante 4y ago
    but training against the entire Internet would still be biased towards English because English is the dominant language used on the Internet.
  - MichaelZuo 4y ago
    > Everyone has to choose what data to train on, you can't train against The Entire Internet, it's a limitless amount of data. Isn't that exactly how OpenAI managed to 10x GPT 3.5 with GPT 4.0?
- famouswaffles 4y ago
  Both of you are kind of misunderstanding a few things. Data used to train the tokenizer is entirely separate from data training the LLM. The tokenizer used to train GPT-3 was old, inefficient and targeted at tokenizing English. That's pretty much all there is to it. It's possible to train a tokenizer that is more efficient and more including of other languages. GPT-4's tokenizer is already far more efficient though still weighted to English.
  - viscanti 4y ago
    > GPT-4's tokenizer is already far more efficient though still weighted to English. Right. It's a general question. Should they be allowed to take the kinds of optimizations they can with tokenization when it's a function of how much data they can use, even if that means some languages get more optimization than others? Or should users of those languages that could be optimized effectively pay a tax out of some sense of fairness?
user_named 4y ago
What makes you think there's a "should"?
- cool_dude85 4y ago
  There's always a should. Society gets a say in what people and corporations can and can't do in (at the very least) the form of laws. There's your should right there.
  - pixl97 4y ago
    Which society? I mean OpanAI is a US company is unsurprisingly going to mostly communicate in English. Are we counting all societies, if so should our software cater to all their demands, language and/or culturally demanded?
famouswaffles 4y ago
Both of you are kind of misunderstanding a few things. Data used to train the tokenizer is entirely separate from data training the LLM. The tokenizer used to train GPT-3 was old, inefficient and targeted at tokenizing English. That's pretty much all there is to it. It's possible to train a tokenizer that is more efficient and more including of other languages. GPT-4's tokenizer is already far more efficient though still weighted to English.

terafo 4y ago

Nope, it's not about dataset. It's just bad tokenizer. Korean has couple of dozen of symbols in it's alphabet. Cyrillic languages have less than 50 symbols in total. Hiragana is 46 symbols. GPT-4 has 32k tokens IIRC. Including most significant alphabets would take less than a thousand.

moelf 4y ago
but the tokenizer is dataset-driven... it tokenize the most common pattern in your dataset to improve efficiency, so it's 100% about dataset?
- sillysaurusx 4y ago
  There's dataset during training, and dataset for the tokenizer. The confusion here is that people are talking about the former, but you're correct that it's the latter. Remember, OpenAI's tokenizer was created in an era when 125MB was considered large for a language model. It's hard to fault them for making something that lasted four or five years.
  - dragonwriter 4y ago
    > Remember, OpenAI’s tokenizer was created in an era when 125MB was considered large for a language model. GPT-2 and GPT-3 have different vocabularies and maximum token #s, which (even if the tokenizer architecture is the same) implies a different tokenizer model. GPT-3.5 might share the GPT-3 tokenizer, but even then I’d expect GPT-4 to have its own. But even if they are using the tokenizer from GPT-3, its not from “an era when 125MB was considered large for a language model”.
    - sillysaurusx 4y ago
      Actually, GPT-3's tokenizer is the same as GPT-2. https://datascience.stackexchange.com/a/109483 https://datascience.stackexchange.com/a/109483 You had me questioning myself for a minute. (The vocab size is still 50257. Even rounded up to a multiple of 128 for better sharding across the vocab embedding, only the first 50257 are used.) Believe it or not, 125M was large at the start of the GPT-2 era. No one knew LLMs could do anything interesting, let alone that they'd change the world.

tlrobinson 4y ago

I think yes, but more precisely the tokens were chosen to optimize training on a dataset that's biased to English content. I am curious how the token set affects quality of responses, ignoring the factors related to token count mentioned in the post (cost, prompt expressivity, latency, etc) Is it always better for the token set to be "native" to the majority of the training dataset and prompts/completions, or is it possible there's some "intermediate representation" (in compiler terms) that would be better?

k8si 4y ago
I don't know what you mean by compiler terms but basically, worse tokenizer = worse LM performance. This is because worse tokenizer means more tokens per sentence so it takes more FLOPs to train on each sentence, on average. So given a fixed training budget, English essentially gets more "learning per token" than other languages.

GaggiX 4y ago

The article only clarifies that the dataset used to train the tokenizer is baised, not the entire dataset used by the GPT model.

famouswaffles 4y ago

Data used to train the tokenizer is entirely separate from data training the LLM. The tokenizer used to train GPT-3 was old, inefficient and targeted at tokenizing English. That's pretty much all there is to it. It's possible to train a tokenizer that is more efficient and more including of other languages. GPT-4's tokenizer is already far more efficient though still weighted to English. You can test it here https://tiktokenizer.vercel.app/ https://tiktokenizer.vercel.app/