Paul Stenhouse: Identifiers on AI images and Google's Gemini
Saturday Morning with Jack Tame · Newstalk ZB
Audio is streamed directly from the publisher (pdst.fm) as published in their RSS feed. Play Podcasts does not host this file. Rights-holders can request removal through the copyright & takedown page.
Show Notes
AI generated images are getting an identifier
Images created by Open AI's DALL-E technology will now come with a new label, a CR in a circle up in the top left corner. That's the mark that’s going to be on images produced by a range of AI companies. The only trouble there is it can pretty easily be cropped out.
Open AI will also be adding metadata to the image file. While it's a nice idea, it's wildly impractical. If you take a screenshot of the image, the metadata disappears. If you post it online, the metadata isn't easily visible, or can even be wiped.
But the same tech actually shows promise for verifying reality. The Coalition for Content Provenance and Authenticity (C2PA) is working with camera companies, and even newsrooms, to embed metadata in real images to verify their authenticity. It will stay as the image is edited and be able to be verified using an online tool with the full history.
You'll be hearing a lot about Gemini
Google's Bard and Duet are being merged under the new name. Gemini will be the name of their large language model and there'll be Pro and Ultra versions. It's part assistant, part chatbot, part search engine — likely the future of Google.
LISTEN ABOVE
See omnystudio.com/listener for privacy information.