You can be compensated for your online posts, provided they are utilized for AI training.
According to Dr. Margaret Mitchell, the chief ethics scientist at Hugging Face, an open-source AI company, there is a pressing need for AI firms to trace AI-generated content back to its original creators.
“Many creators—including artists, writers, and everyday users—are losing out on compensation for their contributions,” she stated during her talk at AI Everything in Cairo, Egypt.
“I envision a future where we can truly identify the sources of input that make AI outputs possible and adequately reward them.”
Generative AI heavily relies on certain creators more than others. Some AI-generated works exhibit distinct links between input and output, such as a recognizable writing style or an artist’s signature.
Recently, renowned Japanese animator and film director Hayao Miyazaki criticized AI-generated images that mimic the unique style of his Studio Ghibli films.
But the issue extends beyond musicians and artists, as large-scale language models (LLMs) like ChatGPT and Google Gemini are trained on extensive online resources.
“We are all creators,” Mitchell emphasized, as reported by BBC Science Focus. It is essential that reward models recognize contributions from all online users, whether it’s a poem or a vacation sunset photo taken five years ago.
Fortunately, there are emerging models that can track the relationship between input and output, rewarding creators based on their contributions.
However, such a system is not yet in place, and existing AI business models hinder the funding required to develop it, Mitchell said, although some AI companies are exploring potential solutions.
For instance, in a document from 2021, AI company Anthropic’s CEO Dario Amodei proposed a “crazy idea” for a reward distribution model akin to the monetization platform Patreon, which was recently opened by court order.
Mitchell noted that existing LLMs could implement known technological strategies to facilitate this model. Clustering algorithms, for example, could help track similarities and attribute authorship.
To maintain user privacy, this model would require consent, allowing users to opt in for their data to be linked to their identity (with compensation) or remain anonymous.
“To foster innovation, we need to pave the way for this kind of research,” Mitchell asserted. “Currently, the path forward is completely closed.”
Read More:
Source: www.sciencefocus.com
