Part 3/9:
Central to Huang's message is that data is the "food of AI"—necessary for learning and operation. As AI agents evolve, they don’t just learn from data; they continually operate on it, making data ingestion and management critical.
He underscores a fundamental shift: whereas traditional enterprise knowledge was stored in structured formats like relational databases with SQL queries, the future calls for semantic AI processing—a paradigm that discovers meaning through embeddings and neural network-based indexing. Unlike conventional hash tables or trees, vectorized databases facilitate fast, neural-driven nearest-neighbor searches across multimodal data types such as videos, texts, PDFs, chemical compounds, and health records.