gpt-3 VS DALLE-mtf

Are generative AI models useful for learning a language, and if so which languages? Over 90% of ChatGPT's training data was in English. The remaining 10% of data was split unevenly between 100+ languages. This suggests that the quality of the outputs will vary from language to language.

I have doubts it was extensively trained on German data. Who knows about GPT4, but GPT3 is ~92% of English and ~1.5% of German, which means it saw more "die, motherfucker, die" than on "die Mutter".

(https://github.com/openai/gpt-3/blob/master/dataset_statisti...)

Catalan was 0.018 % of GPT-3's training corpus. https://github.com/openai/gpt-3/blob/master/dataset_statistics/languages_by_word_count.csv.

>Certainly it is pleasing that they are not just doing Hindi, but some of these languages must be represented online by a very small corpus of text indeed. I wonder how effectively an LLM can be trained on such a small training set for any given language?

as long as it's not the main language it doesn't really matter. Besides English(92.6%), the biggest language by representation (word count) is taken up by french at 1.8%. Most of the languages GPT-3 knows are sitting at <0.2% representation.

https://github.com/openai/gpt-3/blob/master/dataset_statisti...

Competence in the main language will bleed into the rest.

And from skimming their GH page, it seems it'd be costly to host as well

This vision is in line with EleutherAI, a non-profit organization founded in July 2020 by a group of researchers. Driven by the perceived opacity and the challenge of reproducibility in AI, their goal was to create leading open-source language models.

My current idea is to utilize the EleutherAI pythia (Databricks Dolly). I would like to know whether translating the Dolly-15k dataset into the desired language using state-of-the-art translation techniques like DeepL would be a viable approach to fine-tune the Pythia base model. I want to use this model for semantic search, so perfection is not a necessity.

There are open source AI efforts, like EleutherAI. Needless to say, they are lagging behind big players, but it's better than nothing.

The new GPT 3.5 isn't aware what are GPT-3.5 or davinci-002 (repeatable) and claimed that it was designed by EleutherAI and has only 6 bil parameters (wasn't been able to repeat but didn't really try).

Hi, my name is Stella Biderman and I run EleutherAI, the one of the foremost non-profit research institutes in the world that trains and studies large language models. I have been involved with the majority of models to hold the title “largest open source GPT model in the world” and have dabbled in exploring using plagiarism detection tools to identify code written by GPT-J.

dolly-v2-12bis a 12 billion parameter causal language model created by Databricks that is derived from EleutherAI’s Pythia-12b and fine-tuned on a ~15K record instruction corpus generated by Databricks employees and released under a permissive license (CC-BY-SA)

It is true that Emad needs to find an appropriate business model. The good news is that the hype is still undergoing. I'm sure that Emad can grab another round of liquidity injection. He got plenty of resources. Remember he is also from the finance industry. He got https://www.eleuther.ai/ which can supply a secured, in-house custom LLM equivalent to bloombergGPT.

By promoting fully open-source AI, i.e. making datasets, models, methodology and codebases freely available and transparent. What OpenAI claimed to be aiming for, basically.