LAION Withdraws AI Data Sets Following Discovery of Child Exploitation Content
Summary:
The popular artificial intelligence data sets LAION-5B and LAION-400M from LAION, a German non-profit, have been withdrawn following Stanford researchers' discovery of suspected child exploitation material. Though such content doesn't necessarily influence AI model outputs drastically, it can have an impact, especially if repeated. LAION promises strict measures against illegal content and works with organizations to check and ensure link safety in their datasets. To ensure data set safety, LAION has removed the controversial data sets before they are republished.
A popular artificial intelligence data set commonly employed to educate models such as Stable Diffusion and Imagen has been withdrawn by its provider. This comes after a research revealed the presence of numerous suspected instances of child exploitation material. The organization responsible for this data set, LAION - the Large-scale Artificial Intelligence Open Network, is a non-profit from Germany that contributes open-source AI models and training sets for a host of text-to-image applications.
Researchers from the Stanford Internet Observatory’s Cyber Policy Center published a report on December 20 stating that they have uncovered 3,226 instances of suspected child exploitation material within the LAION-5B data set. According to David Thiel, the Cyber Policy Center's Big Data Architect and Chief Technologist, third parties have verified a significant portion of this material. Thiel also stated that existence of such content does not necessarily imply it will drastically modify the results of models trained on the data set; however, it can potentially exert some influence. The repetition of identical exploitation material could be especially problematic due to its reinforcement of images of distinct victims, he added.
LAION introduced the LAION-5B dataset in March 2022, comprising 5.85 billion image-text pairs. The organization maintains strict policy against illegal content and partners with entities like the Internet Watch Foundation among others to examine and confirm the safety of hyperlinks in the LAION datasets. These links are validated with the help of filtering tools established by their community and associated organizations.
LAION, signaling caution, has declared it has removed the questionable data sets, including both LAION-5B and LAION-400M. This step was taken in order to ensure the data sets' safety before they are made available for usage again.
Published At
12/21/2023 9:45:55 AM
Disclaimer: Algoine does not endorse any content or product on this page. Readers should conduct their own research before taking any actions related to the asset, company, or any information in this article and assume full responsibility for their decisions. This article should not be considered as investment advice. Our news is prepared with AI support.
Do you suspect this content may be misleading, incomplete, or inappropriate in any way, requiring modification or removal?
We appreciate your report.