July 2023 Community News!
Data Labeling with GPT-4 in Label Studio
Data quality is the cornerstone of any machine learning project. The traditional way to guarantee data quality is through the time-consuming and expensive task of data labeling. But what if we could leverage the latest advancements in large language models (LLMs) to transition from “data labeling” to “label review”?
Enter GPT-4. With over 100 million users, ChatGPT (built on GPT-4) is one of the most popular state-of-the-art language models available. Its ability to understand and generate human-like text has made waves in the natural language processing (NLP) community.
In this new article, Jimmy Whitaker, our Data Scientist in Residence, shows you how to use GPT-4 with the Label Studio ML backend to automate your text labeling workflow. You’ll learn to shift from the time-intensive task of data labeling to the far more efficient process of reviewing and refining annotations. Check it out to learn more about how you can accelerate your workflow!
This Week! Zero-Shot Machine Learning Workshop
Join our Sr. Developer Community Advocate Erin Mikail Staples, and our Data Scientist in Residence Jimmy Whitaker, for a Zero-Shot Machine Learning Workshop and explore how you can make the most of large language models (LLMs) to speed up dataset curation.
This is a free workshop, taking place virtually on July 26, 2023 at 3PM ET.
You can RSVP here: Zero-Shot Machine Learning Workshop.
Generative AI in Label Studio Survey
Label Studio’s flexible interface and machine learning API make it an excellent choice for emerging generative AI workloads, with templates to support reinforcement learning, LLM retraining, and automatic data annotation, but with a rapidly shifting landscape, the core Label Studio team wants to better understand your needs for integrating Generative AI into your workloads.
To help us along with this, we’ve created a short survey about the present and future of Generative AI. It will only take a few minutes to fill out and will help drive the future development of Label Studio integrations with Large Language Models and other foundation models.
Featured Integration: Hugging Face
Looking for a fast way to launch your annotation project? Label Studio in Hugging Face Spaces makes it possible to deploy your own cloud instance of the Open Source Edition of Label Studio in minutes.
Now Label Studio supports Hugging Face persistent storage! With just a few quick configuration changes, you can transform your Label Studio space into a compact production environment for your annotation and labeling projects.
Read more about how you can Kickstart your Label Studio Annotation Project on Hugging Face Spaces!
Community Shoutouts
Shoutout to our community members and contributors across Label Studio, including Rupert Bedford, Syed, Axel Jacobsen, Guoqiang QI, Alex Fornuto, and Lukas Hennies.
Annotations
- Generative AI and the New Legal Frontier
The MIT Technology Review describes how, in the United States, “it’s becoming increasingly clear that courts, not politicians, will be the first to determine the limits on how AI is developed and used.” With lawsuits piling up, mostly around how publicly available but copyrighted data is used, we could see rapid but lasting changes in the legal framework for AI models. - Generative AI in the MoMA?!
For those debating, “Is Generative AI, Art?” Refik Anadol’s  dares to probe that question further to the reader. The artwork displayed in the MoMA’s lobby is based on a custom machine-learning model trained off of MoMA’s collection to create something entirely new. Curator Michelle Kuo, and Ford scholar-in-residence and art historian Joan Kee discuss what machine learning is to art, as well as Anadol’s piece on the MoMA Blog. - 💅🏻 An AI-Generated Barbenheimer Trailer
ICYMI, someone created an AI-Generated movie trailer for a fictitious Barbenheimer movie. And, for what it's worth, it tells us a lot about the state of Generative AI tooling (and possibly a cross point of our own interests this summer). The trailer feels eerily realistic, yet it was created by an independent team. Check it out for yourself!