Striking the balance

Content quality vs. copyright in the landscape of AI-generated works

In the world of Artificial Intelligence (AI), image generation has seen tremendous growth in recent months, with new players entering the market to challenge the established giants.

A recent comparison of AI-generated images from Adobe's Firefly and the self-funded 11-person team at MidJourney has attracted attention to the differing approaches taken by the two companies. It highlights the fact that the output of AI models is largely dependent on the data curation they are trained on and the algorithms used to develop them.

AI models, like MidJourney, can generate higher-quality outputs by incorporating copyrighted materials in their training data. However, this approach may pose a risk of copyright infringement for companies and does not compensate original creators fairly. Conversely, AI models like Firefly that solely rely on open-license or stock content for training run the risk of producing lower-quality outputs.

If a company prioritizes copyright over image quality, Adobe's Firefly may be a better choice. Nevertheless, the attribution and copyright strategies of both Adobe and MidJourney have limitations and are not ideal for long-term use.

As AI-generated content becomes more widespread, companies must weigh the importance of content quality against copyright safety to determine the optimal approach for their needs.