3 ms·

> Transparency in Model Training: Companies must disclose information about the data used to train AI models, ensuring that the process is transparent and accou

> Transparency in Model Training: Companies must disclose information about the data used to train AI models, ensuring that the process is transparent and accountable. Big win for open source right there. Knock knock, it's the dataset inspector, open up! > Compliance with EU Copyright Rules: Ensuring that AI models respect intellectual property rights. This one's gonna be a rather tall order. Probably means removing all copyrighted material from the training set?

viraptor 2y ago
Ideally it means properly licensing the content not just removing it. It will be interesting to see how the pricing evolves over time.
niemandhier 2y ago
Not necessarily, the regulations for datamining are quite well defined. Could e.g. mean that you demonstrate that you have processes that check if the material you liked was flagged as „do not mine“. https://www.reedsmith.com/en/perspectives/ai-in-entertainment-and-media/2024/02/text-and-data-mining-in-eu https://www.reedsmith.com/en/perspectives/ai-in-entertainmen...
- Findecanor 2y ago
  I interpret the text in the Act and the Directive that "opt-in" is still the default. The "opt-out" clause applies only to research and preservation. Researchers and archivists don't have to ask the rightsholder unless the content is flagged. Commercial users always have to.
  - niemandhier 2y ago
    If I am not mistaken, Article 4 of „the copyright directive in the Eu single market“, indicates an opt out model for text and data mining on copyrighted material in the case for commercial use. https://eur-lex.europa.eu/eli/dir/2019/790/oj#d1e961-92-1 https://eur-lex.europa.eu/eli/dir/2019/790/oj#d1e961-92-1 Can someone point me to the ai act article that would change that ?
    - Findecanor 2y ago
      You're right. The paragraph in the act (105) isn't very well written. You could interpret it as "opt-in required" until you read the referred document or the last sentence.

viraptor 2y ago

Ideally it means properly licensing the content not just removing it. It will be interesting to see how the pricing evolves over time.

niemandhier 2y ago

Not necessarily, the regulations for datamining are quite well defined. Could e.g. mean that you demonstrate that you have processes that check if the material you liked was flagged as „do not mine“. https://www.reedsmith.com/en/perspectives/ai-in-entertainment-and-media/2024/02/text-and-data-mining-in-eu https://www.reedsmith.com/en/perspectives/ai-in-entertainmen...

Findecanor 2y ago
I interpret the text in the Act and the Directive that "opt-in" is still the default. The "opt-out" clause applies only to research and preservation. Researchers and archivists don't have to ask the rightsholder unless the content is flagged. Commercial users always have to.
- niemandhier 2y ago
  If I am not mistaken, Article 4 of „the copyright directive in the Eu single market“, indicates an opt out model for text and data mining on copyrighted material in the case for commercial use. https://eur-lex.europa.eu/eli/dir/2019/790/oj#d1e961-92-1 https://eur-lex.europa.eu/eli/dir/2019/790/oj#d1e961-92-1 Can someone point me to the ai act article that would change that ?
  - Findecanor 2y ago
    You're right. The paragraph in the act (105) isn't very well written. You could interpret it as "opt-in required" until you read the referred document or the last sentence.