AISilicon Canalsabout 2 hours ago
AI may exhaust public text by 2032
7 min read
Epoch AI estimates frontier models could fully use the effective stock of public human text, about 300 trillion tokens, between 2026 and 2032. Developers may then turn to private datasets and synthetic data, with a 2024 Nature paper warning of model collapse from indiscriminate recursive training.
Level
Hype check
Tap to vote and see what everyone thinks.
Summary by ByteBrief
