Back in high school, I attended a seminar on overseas undergraduate education options. The speaker explained that if you simply ask a university for all its materials, your request is likely to be met with silence or delays. However, if you include a token amount—say, $5—with your request for select materials, you’re much more likely to receive a positive response. That small fee acknowledges the inherent work behind those materials and makes clear your genuine intent to learn.
I believe this principle can be effectively applied to AI’s approach to copyrighted texts. As it stands, AI systems are trained on vast amounts of content, including copyrighted works, without a direct economic exchange that respects the value of those texts. Imagine if AI developers adopted a policy akin to the university scenario: for each work ingested, they would include a token fee (for example, the publisher’s non-discounted sale price). This fee would serve as a respectful acknowledgment of the creator’s or publisher’s effort, under the understanding that the usage is analogous to human consumption—carefully moderated to avoid excessive verbatim reproduction, much like TV shows that only use brief spots from commercial cinema.
Such an arrangement would not only compensate the publishers fairly but also reassure them that each instance of use is part of a larger, value-adding ecosystem. It’s a controlled and respectful model that treats AI’s consumption of content like a licensed, single-sale transaction rather than an exploitation of intellectual property.
I call for stakeholders in AI development and content publishing to consider a token-fee model for training on copyrighted texts. This framework—much like the university analogy—could provide a balanced means of advancing technology while honoring and financially supporting the creative works that fuel it, ensuring that this isn’t exploited as a free-for-all but is managed in a manner akin to personal, respectful consumption.