OpenAI offers a way for creators to opt out of AI training data. It's so onerous that one artist called it 'enraging.'

OpenAI has introduced a mechanism for artists to opt out of having their work included in DALL-E 3’s future training data, responding to growing pressure from creators who feel their intellectual property is being exploited without consent. However, the implementation is widely criticized as ineffective, as it requires submitting individual removal requests for each image rather than allowing a single, comprehensive exclusion for an entire portfolio. This bureaucratic hurdle effectively discourages most creators, particularly those with large bodies of work, from participating in the process. The practical utility of this opt-out is further diminished because it applies only to future iterations of the model. Artworks already ingested into the current version of DALL-E 3 cannot be unlearned, meaning the AI retains the conceptual understanding derived from the data even if the images are subsequently removed. This creates a scenario where creators have limited control over the immediate use of their content, highlighting the persistent challenge of retroactively addressing data scraping in AI development. This development is significant for open data discourse as it illustrates the tension between open access and creator rights in the age of generative AI. While OpenAI acknowledges the need for opt-out mechanisms, the design of its process suggests a preference for maximizing data acquisition over respecting individual preferences. It underscores the urgent need for clearer, more equitable standards regarding data consent and ownership, challenging the industry to balance innovation with ethical considerations for original content creators.

Source: businessinsider.com
Published on 2023-09-30

Online space has revolutionised information flow — Right to Information Commission »

Open Data

Related news