The NeuLab Chronicles: A Deep Dive into Multilingual NLP and Machine Translation
In the world of academia, there’s a place where the magic of multilingual natural language processing (NLP) and machine translation (MT) happens. It’s called NeuLab, and it’s led by none other than Graham Neubig, an Associate Professor of Computer Science at Carnegie Mellon University.
NeuLab is a playground for NLP enthusiasts. They’re busy incorporating broad knowledge bases into NLP models and generating code like it’s going out of style.
The Zeno GPT-MT Report: A Tale of Large Language Models and Special-Purpose Machine Translation Models
Graham recently put out a report comparing large language models (LLMs) with special-purpose machine translation models. The contenders? Google Translate, Microsoft Translate, and DeepL. The verdict? GPT-4 can hold its own when translating from English to other languages, but it gets a bit winded with very long sentences.
The Cost of Translation: GPT-3.5 Turbo vs. Google Translate and Microsoft Translator
When it comes to the cost of translation, GPT-3.5 Turbo (the engine behind the free version of ChatGPT) is a bargain compared to Google Translate and Microsoft Translator. However, GPT-4 (available via OpenAI’s subscription) is a bit of a splurge.
Graham predicts that companies will likely pivot towards using general-purpose LLMs and fine-tuning them for specific tasks like translation.
Inspired Cognition: A Startup with a Mission
Graham’s startup, Inspired Cognition, is on a mission to provide tools for building and improving AI systems, especially in text and code generation.
The Future of NLP and the Zeno Report
Graham wraps up the pod with some advice for new graduates in the NLP field and his plans for Zeno and the Zeno report.
