Ideas Why we should 'fight like hell' against Big AI
15 snips
Mar 25, 2026 
Karen Howe, investigative journalist and author of Empire of AI, exposes how Big AI concentrates power and harms labor, data, environment and democracy. She traces OpenAI's outsized influence, uncovers hidden moderation labor, and maps parallels between AI firms and empires. The conversation highlights alternatives: smaller task-focused models and community-driven AI projects.
AI Snips
Chapters
Books
Transcript
Episode notes
Kenyan Content Moderator's Trauma
- Karen Howe recounts Alex from Nairobi who labeled violent and self-harm content for OpenAI while earning between $0.46 and $3.74 an hour.
- Alex developed nightmares and severe distress after eight-hour shifts labeling murders, stabbings, and self-harm to train moderation filters.
AI Companies Function As Empires
- Howe frames Big AI firms as modern empires that claim others' data, exploit labor, monopolize knowledge, and justify actions as a moral imperative.
- She links scale-at-all-costs (10,000x GPT scale 2019–2023) to lowered data quality and intellectual property erosion.
Scale Produces Harmful Model Behavior
- Massive scaling of data and compute causes polluted datasets that embed harmful, addictive, and violent content into models.
- OpenAI's safety filters rely on low-paid moderators and automated methods that degrade over long conversations, enabling harms like grooming.
