SlatorPod

#175 Where DeepL Beats ChatGPT with Graham Neubig

Jul 14, 2023

Graham Neubig, Associate Professor of Computer Science at Carnegie Mellon University, discusses his research on multilingual natural language processing and machine translation. He compares large language models with special-purpose machine translation models, highlighting GPT-3.5 Turbo's cost advantage. He predicts companies will use general-purpose LLMs and fine-tune them for specific tasks. The podcast also covers speech-to-speech machine translation system releases and Graham's startup, Inspired Cognition, focused on improving AI systems.

38:53

Episode guests

Graham Neubig

AI Summary

AI Chapters

Episode notes

Podcast summary created with Snipd AI

Quick takeaways

Companies are likely to use general-purpose language models and fine-tune them for translation, showing promising results compared to special-purpose models.

GPT-3 Turbo is competitive with special-purpose models in English translation, while GPT-4 performs well in certain areas that require context.

Deep dives

The Direction of Translation Systems

Companies interested in building translation systems are likely to move towards using a general-purpose language model and fine-tuning it for translation. This approach has shown promising results in comparison to special-purpose translation models.

Introduction

2min

Journey into Language Technology and Machine Translation

5min

The Importance of Context in Machine Translation

20min

The Reasons Behind Open Sourcing AI Models

2min

Transcription solution Whisper and the Use of LLMs for Low Resource Languages

2min

Challenges of Low-Resource Machine Translation and Synthetic Data Generation

8min

In this week’s SlatorPod, we are joined by Graham Neubig, Associate Professor of Computer Science at Carnegie Mellon University, to discuss his research on multilingual natural language processing (NLP) and machine translation (MT).

Graham discusses the research at Neulab, where they focus on various areas of NLP, including incorporating broad knowledge bases into NLP models and code generation.

Graham expands on his Zeno GPT-MT Report comparing large language models (LLMs) with special-purpose machine translation models like Google Translate, Microsoft Translate, and DeepL. He revealed that GPT-4 was competitive from English to other languages, but struggled with very long sentences.

When it comes to cost comparison, Graham highlights that GPT-3.5 Turbo (the model behind the free version of ChatGPT) is significantly cheaper than Google Translate and Microsoft Translator, but GPT-4 (available via OpenAI’s subscription) is more expensive.

Graham predicts that companies will likely move towards using general-purpose LLMs and fine-tuning them for specific tasks like translation. The discussion also covers the recent flurry of speech-to-speech machine translation system releases.

Graham talks about his startup, Inspired Cognition, which aims to provide tools for building and improving AI systems, particularly in text and code generation. Graham concludes the pod with advice for new graduates in the NLP field and his plans for Zeno and the Zeno report.

Remember Everything You Learn from Podcasts

Save insights instantly, chat with episodes, and build lasting knowledge - all powered by AI.

SlatorPod

#175 Where DeepL Beats ChatGPT with Graham Neubig

Episode guests

Podcast summary created with Snipd AI

Quick takeaways

Deep dives

The Direction of Translation Systems

The Language Technologies Institute at Carnegie Mellon University

Graham's Interest in Machine Translation

Evaluation of Large Language Models in Translation

Remember Everything You Learn from Podcasts