AI has run out of training data, warns data chief
Experts warn of ‘crisis point’ ahead for AI data
- Goldman Sachs' chief data officer, Neema Raphael, says that artificial intelligence models are running out of training data.
- This scarcity of data could lead to a “creative plateau” in AI development, with models increasingly relying on synthetically generated data.
- Industry experts, including OpenAI co-founder Ilya Sutskever, have warned of “peak data”, comparing human-generated content to finite resources.
- A Nature article predicted that a “crisis point” for AI training data would be reached by 2028, as the internet's vast information is being depleted.
- The lack of new data may prompt a shift in AI development from large language models to more autonomous 'agentic artificial intelligence' systems.