AI Lab

Blue Sky Innovations AI Lab

We build and test applied AI for media work: agents that research, monitor and publish; simulations for training; and tools that check what's real. Working with partners like NVIDIA, we put AI into real workflows to learn where it helps, where it fails and where human judgment has to stay in charge.

  • AI agents
  • Simulation & training
  • Computer vision
  • Synthetic-media detection
Start a project

A changing media landscape

How we interact with information is changing, and how we create and tell stories is changing with it. AI can research and synthesize information, parse enormous datasets, generate convincing images and video, and increasingly act across entire workflows through autonomous agents. At the same time, audiences are encountering information through AI summaries, conversational interfaces, immersive video experiences, and an unprecedented volume of generated content. These changes create enormous opportunities for communicators and storytellers, while raising equally important questions about trust, authorship, accuracy and human judgment. Hussman faculty and students have an opportunity not simply to respond to these changes, but to explore them as they happen and help shape what professional practice becomes.

What we work with

  • AI Agents

    Research, monitor, analyze and act across workflows, with people brought in when judgment or approval is required.

  • Generated Content

    Generative image, video, audio, graphics and text for ideation, production and creative expression.

  • Computer Vision

    Pose tracking, video analysis and synthetic-media detection that turn images and footage into data for analysis, verification and storytelling.

  • Data Gathering + Synthesis

    Systems that monitor information, search records and datasets, retrieve sources, identify patterns and make complex information accessible.

Projects

Silent screen recording. UNC Detective scores a video of AI-generated Titanic interviews as 81% likely AI (high risk), then scores real MLB highlights as 5% (low risk), with a frame-by-frame timeline for each.

UNC Detective

AI-generated video detection

With our partner NVIDIA

UNC Detective checks whether a video contains AI-generated or manipulated content. Paste a link from YouTube, Facebook or Instagram, or upload a file, and it analyzes every frame, gives the video an overall AI detection score and marks which parts of the timeline look generated. Each result can be shared or downloaded as a report. It was built by researchers at UNC Hussman and runs on NVIDIA AI technology.

NVIDIA AI for Media: Sports Body Pose. At a UNC home game, a player's free throw is tracked with a 3D skeleton while charts follow his joint angles.

Sports Body Pose with NVIDIA

3D body pose tracking for sports media

With our partner NVIDIA

Built with NVIDIA's AI for Media tools and shown at the NAB Show, this project tracks a player's body in 3D from ordinary basketball game footage. It charts pelvis height, knee flex, elbow bend and elbow speed through each phase of a shot, such as the follow-through, turning broadcast video into data for analysis and storytelling.

Participants at a conference table watch a presentation on a large screen during a CISLM training session

AI for Local Journalism with CISLM

AI education and tools for local journalists

Local journalists are covering more ground than ever. With our partner, the UNC Center for Innovation and Sustainability in Local Media (CISLM), we develop AI education and tools that help community and local reporters keep pace with a fast-moving news cycle as their coverage areas grow.

The Rameses dashboard: usage telemetry, a ram avatar, a monthly website action board for UNC Media Hub and a chat panel

Rameses

An AI assistant for the work behind the story

Rameses is an AI assistant built for media and journalism work. Students and journalists use it through a dashboard, voice commands and Discord to research unfamiliar topics, check the UNC Media Hub website for problems, prepare stories for publication and turn public-meeting recordings into searchable, timestamped transcripts. It runs on Hermes, an open-source agent harness from Nous Research and NVIDIA. Journalists remain responsible for verifying its work and making editorial decisions.

More about Rameses

Start a project with the AI Lab

Tell us what you are working on and we will get back to you.

Contact us