Many science volunteering projects train AI models on volunteer annotations, but this is typically a one-way process: volunteers annotate data, and the AI learns from those annotations. We want to explore how models can help you make annotations.
For example, our first tool is a paintbrush for drawing on galaxy images. Our models can suggest where to draw, and you can accept or correct those suggestions.
We know volunteers can do much more than answer questions. You want to dive into the data. You want to find the head-scratching anomalies where scientists go "Wow - I didn't know that could happen!". You want to organise and describe the data in new ways. We hope Experiment can give you tools for all of this.
The Zooniverse team are our collaborators, and with their help we support using Zooniverse log-in and the Galaxy Zoo Talk forum. Together, we hope this new way of building projects will be fun and useful for the Zooniverse community.
The Zooniverse brings together hundreds of projects and millions of volunteers, and so any new tools need to be carefully designed and validated to ensure they work reliably for all projects and volunteers. In contrast, with Experiment, anything can break at any time. The trade-off is that here we can try new things, rapidly, get new types of science done, and (hopefully) inspire new tools that can be deployed across the Zooniverse.
The Experiment platform was built by Conor Klamann, Matthieu Le Lain, and Mike Walmsley. We are grateful for initial funding from the University of Toronto's Data Science Institute. The tools and projects are created by a bunch of astronomers and computer scientists hoping to answer the question: if we could re-invent science volunteering with AI in mind, what would it look like?
Please click here for the Terms of Service and Privacy Policy.