Episode 27: Tracey Jaquith - TV News Archive + AI
Manage episode 409924631 series 3448018
Tracey Jaquith, Founding Coder of the Internet Archive and co-founder of its TV/News archive, will give us a personal tour of it's ChatGPT/GPT4 implementation and dive into its intricacies.
We will get a sneak peek behind the scenes of the TV/News Archive's ChatGPT implementation. "Chyrons," the short textual overlays at the bottom of news, get OCR scanned on four news channels. GPT4 summarizes the chyrons hourly, cleaning up scanning errors and eliminating ads. Results are posted to Mastodon with an API for researchers. The AI inputs and techniques have been setup to avoid hallucinations.
About the Internet Archive's TV/News Archive:
The Internet Archive's TV/News archive (https://archive.org/tv) indexes over 50 stations in over 30 countries, and provides many different ways to look for specific news stories and video clips and share the information with others. These tools may prove indispensable this year towards verifying the authenticity of news clips, for confirming, corroborating and triangulating the original sources of news stories, and for cross-referencing with partner fact checking sites.
About Tracey Jaquith:
Tracey was the founding coder and system architect for Internet Archive in 1996. Recently, she rewrote Archive’s TV recording system to an opensource single server system, capable of 75 simultaneous 24×7 channels, made the TV site “full stack”, and brought the archive.org website to “version 2”. Tracey currently focuses on devops, video, audio, and AR/VR/3D. She is leading the Archive’s conversion to docker containers.
Tracey holds degrees in computer science from Cornell University where she focused on machine vision and robotics. Outside of work, she bikes trails daily. Find her at poohbot.com (https://poohbot.com/), talks (https://tracey.archive.org), Twitter @tracey_pooh (https://twitter.com/tracey_pooh).
FOLLOW Mindplex Podcast:
WEBSITE → https://magazine.mindplex.ai/
Mindplex News → https://magazine.mindplex.ai/news/
Mindplex TWITTER → https://twitter.com/Mindplex_ai/
FACEBOOK → https://www.facebook.com/MindplexAI/
YOUTUBE → https://www.youtube.com/@mindplexpodcast/
Spotify → https://open.spotify.com/show/3RbGNlj9SX8rYOVblnVnGg/
38 episodes