AI development is facing a declining supply of human-generated data for training, as Elon Musk noted in a recent interview with Stagwell CEO Mark Penn.

Musk believes that by the end of 2022, most human knowledge available for AI training had already been exhausted.

While companies like Google, Meta, and OpenAI increasingly rely on synthetic data, Musk has warned about potential risks, including errors or “hallucinations” in AI outputs.

A study by Epoch AI predicts that publicly accessible human-generated data could be depleted between 2028 and 2032, posing a significant challenge for future AI advancements.