3 ms·
Given that big tech has been scraping everything ever written to train LLMs, are there specialized prompts to trick models into spitting out copyrighted works ?
by phtrivier 1y ago
Given that big tech has been scraping everything ever written to train LLMs, are there specialized prompts to trick models into spitting out copyrighted works ?
- baq 1y agoFoundation LLMs are lossy compressed databases, you might never get the exact work back from it.
- Nemo_bis 1y agoYes, if you believe the New York Times: https://www.techdirt.com/2023/12/28/the-ny-times-lawsuit-against-openai-would-open-up-the-ny-times-to-all-sorts-of-lawsuits-should-it-win/ https://www.techdirt.com/2023/12/28/the-ny-times-lawsuit-aga...