"We need to build an economy where the data contributors are paid." - Jaron Lanier, Computer philosophy writer and scientist
A flawed government proposal
The Department for Promotion of Industry and Internal Trade (DPIIT) of India has released a working paper on generative AI and copyright. It suggests a mandatory blanket license fee that AI companies must pay for using copyrighted materials. A new body called the copyright royalties collective for AI training would collect these fees from developers. The goal is to distribute this revenue to registered creators through existing collective management organisations.
AI companies stay safe
This is not a tax on tech giants, because the companies actually win. For a small percentage of revenue, they eliminate significant legal risks regarding copyright infringement. Since their revenue comes from subscriptions, they will likely pass this cost on to consumers. This arrangement protects their business model and profitability while providing them with legal certainty against lawsuits. So this is a problem!
Check out all our posts on National AI policies; click here
Huge funds ending up in a black box!
The clear winners in this setup are the intermediaries. Management organisations typically retain a significant portion of collected money for administrative expenses. Historical data suggests that these schemes often result in revenue ending up in a black box. Funds where specific artists cannot be identified are often disposed of at the discretion of the organisation rather than reaching the creators.
Tokenised problem
The unit of value in AI is a token, which is a fragment of a word or image. Unlike music streaming where plays are countable, AI training disperses a work across billions of parameters. It is statistically impossible to trace a specific output back to a single token or an individual artist. This makes accurate attribution and fair payment distribution nearly impossible.
Small creators, sorry
This proposal is a transfer of wealth from small creators to established artists and intermediaries. Since attribution is difficult, revenue will likely go to superstars who are already overrepresented in datasets. The average individual creator will receive little to nothing, leaving them worse off than before. The entire proposal needs to be relooked at.
Summary
The government proposes a blanket license for AI training to compensate artists, but the plan is flawed. It benefits AI companies by removing legal risk and enriches intermediaries. Due to the technical impossibility of tracking specific data tokens, small creators will likely see no money, making it a wealth transfer to the rich.
Food for thought
If AI learns from statistical mixtures rather than copying specific works, is the concept of copyright obsolete in the age of machine learning?
Check out all our posts on National AI policies; click here
AI concept to learn: Tokens
A token is the fundamental unit of information that large language models process to understand and generate text. It acts as a small fragment of a word or character sequence rather than a complete word. This granular approach allows the model to predict patterns and relationships within vast amounts of data efficiently.
[The Billion Hopes Research Team shares the latest AI updates for learning and awareness. Various sources are used. All copyrights acknowledged. This is not a professional, financial, personal or medical advice. Please consult domain experts before making decisions. Feedback welcome!]
COMMENTS