Academic conferences

Academic conferences can use Studios to enable authors to share fully reproducible work as Studios.

Set up a conference

Follow these steps to set up a conference page on Lightning AI. Please get in touch so our team can help.

Create an organization

Create a Lightning AI organization with the conference name. This is where you'll group all accepted papers and Studios into an easy to search library where anyone can duplicate published Studios and verify results themselves.

Set up the conference library

Our team will work with you to set up a library that organizes and makes public all Studios published by conference publishers.

Authors submit Studios

During the review process, our suggestion is to encourage paper publishers to submit a Studio with their publication review. This will allow reviewers to validate claims made on the papers, inspect code, results and more. Once a paper has been accepted, conference administrators can choose to feature the published Studios in their conference library for the global AI community to explore and reproduce.

Fund authors

Lightning AI gives free credits to all users to empower more AI to become open source. Conferences can provide additional funding for Studio publishers to publish work that might require very expensive GPUs to replicate.

Advantages

Full reproducibility

Academia suffers from a reproducibility crisis, where claims made in paper cannot be 100% verified. Studios allow reviewers to duplicate the author's work in <1 minute by having the author published a Studio which is fully self-contained with all code, files, dependency and environment details. Lightning AI can work with conferences to reduce the cost of doing this.

Transparency

The scientific community can validate work by duplicating a Studio, inspecting the code and results to verify claims. This dramatically increases transparency and accountability at a time when more and more people are getting into AI.

Accelerate reusability

Published work today sits on GitHub repos that people cannot fully reproduce because they lack all environment details, the type of compute required, the right cuda version, and so on.

Published Studios can instantly be duplicated and integrated as a module into follow up work by anyone in the world.

Faster review process

It is much easier to validate claims by duplicating a Studio in a few minutes instead of spending days or weeks debating the merits of a claim, (or even if a claim at all can hold). Many science disciplines that use AI show empirical results as a means to prove claims. The tooling must evolve to enable this new kind of validation. Studios unlock this phase of scientific transparency.