Artificial Image Incident Database

Deepfakes have the capacity to improve storytelling and art, but they can also be applied to produce harms to our social and political systems. The purpose of this project is to help the public avoid harms to the public discourse by people manipulating videos and images to present false or distorting stories. Essentially, this project is to build an index of deepfakes and cheepfakes affecting the public interest (politicians, world events, etc.).

This project builds on top of the codebase of the AI Incident Database, which includes 1700+ public records displayed within an instant search user interface. The project has earned media attention in Wired, VentureBeat, and Vice News and now has a staff of 5 developing the open source codebase. This project builds on top of the open source codebase as a proof of concept that may extend the activities of the sponsoring organization, the Responsible AI Collaborative.

Students engaging in this project should have an entrepreneurial mindset and solve the problems necessary for a public launch at the end of the Capstone. Therefore, students will be expected to take the existing codebase that focuses on AI incidents and make the minimum changes necessary for that codebase to be applicable to the rich media use case (i.e., produce a minimum viable product).

Objectives

  • Open pull requests against the AI Incident Database codebase to make it a repurposeable codebase for incident reporting, rather than a codebase applied exclusively to AI incidents (this should be a lighter deliverable)
  • Extend the image and video hosting capabilities of the AI Incident Database so that it can more effectively index deepfake images and video (this should be a heavy deliverable)
  • Develop the deepfake ingestion criteria and process that will define how/why deepfakes are added to the database
  • Continuously deliver (as in, don't wait until the end of the project to deliver everything) code changes in the form of pull requests that are merged into the main branch of the AI Incident Database
  • Write a blog post for the experimental launch of the image database. The Responsible AI Collaborative will judge reception of the minimum viable product and potentially staff it on an ongoing basis

Motivations

A lie gets halfway around the world before the truth has a chance to get its pants on.

Winston Churchill

The purpose of this project is to help speed the truth along.

Qualifications

Minimum Qualifications:
  • Experience in web development
  • Experience with open source software development (git flow)

Preferred Qualifications:
  • React development experience
  • Cloudinary experience
  • MongoDB experience
  • Web API experience
  • Docker experience


Details

Project Partner:

Sean McGregor

NDA/IPA:

NDA/IP Required

Number Groups:

1

Project Status:

Accepting Applicants

Website:
https://incidentdatabase.ai/
Keywords:
mongodbReactweb developmentDatabasepoliticsDeepfakesFact CheckingCloudinaryAI Incident Database
Card Image Capstone