Revolutionary Certification to Assess Ethical Training of Generative AI Systems

Revolutionary Certification to Assess Ethical Training of Generative AI Systems

A new non-profit organization‌ called​ Fairly Trained is offering​ certifications⁤ to companies⁤ that​ train their⁣ generative AI models on “consented” ⁢data, amid disputes between creators and IP holders and​ generative AI companies over data usage protocols.

Fairly Trained‍ aims‍ to certify AI companies that do not use copyrighted ‍work without a license, emphasizing the⁤ importance of creator consent in training generative AI⁣ systems.

CEO Ed-Newton Rex, a former employee of Stability AI, leads Fairly Trained, with advisers ⁣including Tom Gruber, ⁢co-founder and CTO of Siri, and‍ Maria Pallante, president and CEO of the Association of American Publishers.

The organization currently offers ⁢a single certification, the Licensed Model Certification, which requires companies to meet specific ‌prerequisites for ⁤training data and obtain consent from data providers.

Companies seeking certification must have a robust data due diligence process and maintain records of the training data used for each model training.

For⁤ more information, visit www.infoworld.com

2024-01-24 03:00:05
Source from www.infoworld.com

Exit mobile version