AI companies have effectively consumed nearly all available music data from the internet, according to former OpenAI Co-Founder and Chief Scientist Ilya Sutskever. Speaking at the NeurIPS conference in Vancouver, he declared that pre-training AI on internet data will soon end because "we've achieved peak data...there's only one internet."

While companies like Anthropic argue that training AI on copyrighted content qualifies as 'fair use,' major music industry players strongly disagree. This contentious issue will likely face significant legal challenges in the coming years.

The future of AI is shifting away from large language models (LLMs) that simply pattern-match existing data. Instead, the focus is moving toward:

• 'Agentic' AI systems that can autonomously perform tasks and make decisions
• Models that can reason through problems step-by-step, similar to human thinking
• Development of spatial intelligence and 3D understanding

Stanford's Fei-Fei Li emphasizes that relying solely on 2D internet data is like "building an AI for a flat earth." The next frontier involves teaching AI to reason more like humans rather than just reproducing patterns from existing data.

This development has significant implications for the music industry, as AI companies have already scraped vast amounts of copyrighted content, including music, movies, and books, from both legitimate and piracy websites to train their models.

Related Articles

Previous Articles