AI data scraping hypocrisy: Big tech gatekeeps models after training on public web
OpenAI, Meta profit from billions of scraped pages while charging premium for access...
Deep Dive
The irony? They cry existential threat when they were the ones who made us feel that way first.
Key Points
- OpenAI, Meta, and Google trained on trillions of tokens scraped from the public web without explicit permission
- Access to top models like GPT-4 costs $20–$200/month, while free tiers are being progressively restricted
- Unofficial distillation projects have created 10x cheaper models, sparking legal and ethical disputes
Why It Matters
The data-access double standard threatens trust in AI and may accelerate calls for regulatory oversight.