Tag: training-data

Showing 1-2 of 2

Jun 23, 20267 min readModel Release BriefIntermediate

Outsourced annotators report using public chatbots to generate 'human' training dialogues

Whistleblowers say subcontracted annotators often prompt public chatbots (e.g. ChatGPT) to produce 'human' training dialogues—risking model feedback loops. Practical checklist included.

training-data data-quality crowdwork model-risk ai-safety

+2 more

data-governance newscientist

Model BreakdownsUnited States

Open

Mar 20, 20267 min readFounder NotesIntermediate

Improv Actors Hired to Record Emotional Performances for AI Training — What Performers and Employers Should Know

Job listings seek improv actors to record emotional performances for reusable AI training datasets. Learn licensing risks, work impacts, and steps performers and managers should take.

AI jobs training-data creative-workers ethics

+2 more

policy founder-notes