AI Companies Hiring Actors During Hollywood Strikes to Train AI Raises Concerns Over Data Usage
• AI companies hire out-of-work actors to train AI with their expressions, voices, and movements.
• Previous attempts to generate human avatars have relied on scraped data from the internet without consent.
• Companies like Realeyes and Meta are hiring actors, yet it is uncertain how the collected data will be used and if actors’ likenesses will be commercially reproduced.
• Actors are granted certain rights to their identities, allowing future research using their data and the creation of derivative works.
• Generative AI has had repercussions in the entertainment industry, with deep fakes and de-aging techniques raising ethical concerns.
• As the demand for data increases, participants must understand the potential consequences of their involvement.
Out-of-work actors are finding a new avenue of employment in AI job postings that require their acting skills. These “trainers” teach AI to appear more human by lending their expressions, voices, and movements as data points. However, the implications of how their likeness and performances will be used still need to be determined.
Previous attempts to generate human avatars using facial recognition, biometric analysis, and generative AI models have relied on scraped data from the internet, often without consent. The need for higher-quality data has compelled companies like Realeyes and Meta to hire actors for their projects.
While these companies claim not to reproduce individual likenesses for future projects, it is uncertain how the collected data will be utilized. The job postings promise payment but state that the actors’ likenesses will not be used commercially. Yet, the broader implications of their participation make it difficult for actors to fully understand what rights they may surrender.
The agreement provided to the actors grants Meta certain rights to their identities, allows future research using their data, and permits the creation of derivative works. However, there are limitations in place to prevent machine learning models from generating specific faces or voices for Meta products.
Generative AI has already had repercussions in the entertainment industry, with deep fakes and de-aging techniques raising ethical concerns. Non-A-list celebrities, including background actors, have been subjected to digital body scans without payment, fearing they will be excluded from future opportunities.
As the demand for data to power generative AI continues, participants must understand the potential consequences.