Weekly AI and NLP News — February 27th 2023

OpenAI Foundry, Hugging Face with AWS, and Roblox

Fabio Chiusano
February 27, 2023

Here are your weekly articles, guides, and news about NLP and AI chosen for you by NLPlanet!

OpenAI Foundry will let customers buy dedicated compute to run GPT3 and their other models. It will soon be possible to have dedicated GPT3 models.
Hugging Face and AWS partner to make AI more accessible. Making Generative AI models more open-source.
How AI Can Help Create and Optimize Drugs To Treat Opioid Addiction. It could save thousands of lives every year.
Roblox Is Bringing Generative AI to Its Gaming Universe. The company aims to draw on the new technology’s code-writing ability to make its digital worlds even more customizable.
How chatbots can change journalism. Or not. Experiments refining newspaper articles with Claude, the chatbot from Anthropic.
Google Research on Natural sciences. Advances in physics, biology, and natural sciences.

MIT course on Introduction to Data-Centric AI. A free and open course that focuses on data quality.
Lessons learned while using ChatGPT in education. Experiments from a teacher giving homeworks with ChatGPT.
Writing Essays With AI: A Guide. Advice on how to incorporate generative AI in your writing process.
Text-to-Image Diffusion Models: A Guide for Non-Technical Readers. A simple explanation of how text-conditioned diffusion models work.
Overcoming The Limitations Of Large Language Models. Ideas to complement the intelligence of LLMs.

Pretraining Language Models with Human Preferences. Exploring alternative objectives for pretraining LMs in a way that also guides them to generate text aligned with human preferences.
How Good Are GPT Models at Machine Translation? A Comprehensive Evaluation. GPT models achieve very competitive translation quality for high resource languages, while having limited capabilities for low resource languages.
The Wisdom of Hindsight Makes Language Models Better Instruction Followers. An alternative approach to RLHF: converting feedback to instruction by relabeling the original one and training the model for better alignment in a supervised manner.
Scaling Vision Transformers to 22 Billion Parameters. From 4B to 22B parameters, ViT demonstrates increasing performance with scale.
Zero-Shot Information Extraction via Chatting with ChatGPT. Using ChatGPT for entity-relation triple extraction, named entity recognition, and event extraction.
Aligning Text-to-Image Models using Human Feedback. RLHF for text-to-image models.
LLaMA: A repository for Open and Efficient Foundation Language Models. A collection of foundation language models ranging from 7B to 65B parameters by Meta.

Thank you for reading! If you want to learn more about NLP, remember to follow NLPlanet. You can find us on LinkedIn, Twitter, Medium, and our Discord server!