09: "Fairly Trained" datasets!
A few months ago, OpenAI the company behind ChatGPT, reportedly asked the British Parliament to authorize the free use of copyrighted works. Their argument: It is impossible to train artificial intelligence without this type of data. The statement sounds like a free pass for copyright infringements until the law clarifies what is really allowed and what is not. The topic is particularly debated in the music industry. This is because pieces of music are used in large datasets to create new songs. Ed Newton-Rex did not want to accept this. Last year, the American left Stability AI - whose AI can be used to generate music among other things - because too much was being stolen. He has now launched “Fairly Trained” to counteract this. We talked about this. Also Diaa El All will join later. He is the Co-Founder of Soundflul which is a generative AI company that uses the service of Fairly Trained.
Any feedback? Contact me on my homepage:
Any feedback? Contact me on my homepage:
I produced all of "The Illiac Suite" myself (with a little help from A.I.):
Voices: ElevenLabs
Image: Dall-E 2
Voices: ElevenLabs
Image: Dall-E 2
Music of the episode:
Thanks for listening!
...and check out my agency:
DISCLAIMER: The content displayed in this podcast, including but not limited to the musical renditions and AI voice synthesis, is intended solely for entertainment.
Creators and Guests
Host
Dennis Kastrup
Dennis is a radio journalist in the music business since over 20 years. He has conducted over 1000 interviews with artists from all over the world and works for major public radio stations in Germany and Canada. His focus these days is on “music and technology” – Artficial Intelligence, Robotics, Wearables, VR/AR, Prosthetics and so on. He produces the podcast “The Illiac Suite - Music And Artificial Intelligence”. This interest made him also start „Wicked Artists“: a booking agency for creative tech and new media art.
