Random Thoughts and Observations on Data Science and Beyond

Who is responsible for Generative AI: A first step…


For many millennia the author of a work of art wasn’t acknowledged.

If you draw like Leonarda da Vinci can you rightly say that your skills are like the great man. If you use Generative AI to paint like Vincent van Gogh and Leonardo Da Vinci combined in one painting would the answer change. Can authorship be claimed for AI Generated great works of art? Does the same apply for text too? The promise and the expected rise of Generative AI means these questions need to be addressed sooner rather than later.

To my mind: the responsibility of the artifact (especially true when the artifact is text, video,…), the economic interest, and the right to be identified as the creator of the artifact may have to be dealt with separately in the world of Generative AI.

Responsibility: When ChatGPT is used to write a report the person who shares the report ought to be responsible for the content of the report. And this is only fair as it can be reasonably expected that the person sharing that report agrees with its contents. At times this could be less relevant for fine arts but is applicable here too.

Economic Interest: For paintings on canvas: the natural constraint to create multiple originals was truly a natural constraint. In the digital world there isn’t such a natural constraint in creating multiple originals. So, to create an artificial scarcity by making any text or image non-fungible —like in the case on Non Fungible Tokens (NFTs) used for digital forms of art — seems regressive to say the least unless the economics of scale in the form of a nominal or low price is given to the consumer.

Right to be identified as the creator: right to be the creator A new norm would be required to say how economic interests and rights to be assigned as the creator be split between the user and the creator of Generative AI models. This could get linked to the economic interests too.

While there are some parallels from the publishing industry, Generative AI will need to evolve its own practice because the contribution from the Generative AI model is of a very different nature. Such a practice could take some time to develop. A consensus on how to address responsibility could be an important first step.

Aniruddha M Godbole is an inter-disciplinary expert. He is a continuous learner. These are his personal views.