LAION-5B dataset

Transparency and accountability 🙈

Considered significant for its scale and relative openness, LAION-5B is seen to suffer from several notable transparency and accountability limitations: 

• Data provenance. There is a lack of clear documentation of where specific images originated, hindering independent verification and auditing efforts, making it difficult for researchers to assess the ethical implications of using the dataset[2][4]. The obfuscation of information surrounding dataset composition limits transparency and prevents the establishment of regulatory guidelines necessary for responsible AI development.

• Data collection. Detailed documentation about the processes used to collect, filter and curate the images in the dataset, hindering the ability to critically assess the dataset’s quality and the ethical considerations involved in its creation. Furthermore, the absence of a "human in the loop" during data collection means that there is minimal accountability for content included in the dataset - a reliance on automated systems that has resulted in a dataset that may perpetuate biases and include harmful material without proper checks or balances.

• Data integrity. Information about the methods and criteria used for cleaning, annotating, and pre-processing the images, making it challenging to assess the dataset's reliability and integrity.

• Privacy. Information on whether or not explicit consent was obtained from the individuals depicted or the content creators, raising transparency issues regarding the permissions and rights associated with the images.

Risks and harms 🛑

Critics accuse the LAION-5B dataset of violating copyright and privacy, perpetuating bias and discrimination, enabling misinformation and disinformation and other harmful content, including the exploitation and abuse of children.

More broadly, it is seen to have contributed to the degradation of creativity and the loss of jobs.

Legal, regulatory 👩🏼‍⚖️

• Robert Kneschke v. LAION

• Andersen et al v. Stability AI Ltd. et al (pdf)

Research, advocacy 🧮

• Knowing Machines (2024). Models all the way down

• Carlini N. et al (2023). Poisoning Web-Scale Training Datasets is Practical

• Thiel D., Hancock J. (2023). Identifying and Eliminating CSAM in Generative ML Training Data and Models

News, commentary, analysis 🗞️

• https://hbr.org/2023/04/generative-ai-has-an-intellectual-property-problem

• https://www.techpolicy.press/laion5b-stable-diffusion-and-the-original-sin-of-generative-ai/

• https://openfuture.eu/note/seeing-like-an-algorithm-a-closer-look-at-laion-5b/

• https://www.newyorker.com/culture/infinite-scroll/is-ai-art-stealing-from-artists