| | |

AI Growth through Expert Communities

In the creator economy, the creative individual sells content. The more attention the content captures, the more valuable it is. The incentive for the creator is status and payment for consumption of their content. Distribution channels are Internet platforms, where content is delivered as intended by the author, the platform does not transform it (other than to distribute it more easily, or ensure it meets terms and conditions of use). 

What if AI built with a conversational interface, is the distribution channel? Who should get paid for content that AI produces?

AI software, built around a large language model (LLM), makes content on the basis of patterns it learns from training data, user context, and user inputs. If the author’s content is distributed exactly as it is in training data, then there is no need to use AI – the distribution channel simply carries content as originally created.

But if the distribution channel is in fact only using content to learn patterns, then content from many authors will influence what is learned, and how that combines with contextual data about a user, and user’s input when interacting with the AI system.

When an LLM-based AI is the distribution channel, the author becomes less important. There is no obvious mechanism to build an author’s status, or to monetize their content. If all you could listen on Spotify were remixes that Spotify (the software) made from the artist’s original work, it would be harder to attribute quality content to individual artists.

A way to try to preserve incentives to individual authors could be to define measures of the contribution parts of training data have, in outputs produced by AI. For example, I ask the AI a question, it computes the answer, and shows me the list of names of people who contributed to training data that were used to provide the answer to me. 

In many cases, however, the answer that AI will provide will have been learned from training data that reflects knowledge of many experts, developed and validated by may others, over a long time. Every question which requires expert knowledge grounded in scientific research, will be hard to trace to individual authors/experts. Instead, it will point to what Lisa Herzog calls an epistemic community in her book “Citizen Knowledge“. 

“Given the importance of trustworthy collaboration for the generation and verification of many forms of expert knowledge, it is a genuine question whether such knowledge can be said to be held by individuals rather than by the communities in question.143 Many historians of science have underlined the curious fact that many inventions have been made almost simultaneously, but independently, by several scientists.144 This points to the fact that epistemic communities develop knowledge together, so that achievements by individual members are less important than what the group achieves as a whole. Moreover, economic historians have shown that the technical innovations and the economic growth of the last centuries were, to a great extent, due to the interaction of various forms of knowledge by various individuals, who together were able to come up with innovations and break-through discoveries.”

Section III.4 in Lisa Herzog, Citizen Knowledge, Oxford University Press, 2024

To stimulate growth, companies making AI need to consider whether they can grow quality content creation through the creator economy, the compensation of individual authors, or an epistemic community economy, where incentives return to communities, to stimulate further knowledge creation, through the preservation and growth of the community.

Similar Posts