‘Data poisoning’: How artists are fighting back against Artificial Intelligence image generators

Data poisoning, exemplified by tools like Nightshade, allows artists to subtly alter images so that unauthorized AI training sets learn incorrect associations. This defensive tactic causes generative models to produce unpredictable and distorted outputs when processing related concepts. The primary conclusion is that indiscriminate scraping of online data undermines the integrity of AI systems and violates the moral rights of creators who lack control over their work. This phenomenon highlights a critical tension in open data practices, where the assumption that publicly available information can be freely repurposed conflicts with the need for ethical sourcing. The article suggests that relying solely on technical fixes, such as ensemble modeling or audits, is insufficient. Instead, it advocates for stricter governance regarding data provenance, emphasizing that consent and licensing must precede the inclusion of creative works in training datasets. The relevance to open data lies in the urgent need to redefine data collection standards. Just as facial recognition systems faced backlash for privacy violations, generative AI faces scrutiny for intellectual property infringement. This article argues that open data initiatives must evolve to respect creators’ rights, shifting from a culture of unrestricted harvesting to one that prioritizes authorized, transparent, and respectful data usage to ensure both technological reliability and ethical compliance.

Source: scroll.in
Published on 2023-12-22

'Hope this helps': Emails reveal police had detailed information about Nathaniel Train almost a year before Wieambilla terror attack »

Open Data

Related news