Embeddings as the next unique identifier: from IP addresses to every word in context

StewartalsopIII · x · 2026-09-10

Investor Stewart Alsop III offers a computing-history observation from building apps: computers' core trick is giving every tiny piece of data a unique identifier—first spreadsheet rows, then IP addresses, then DNS names.

LLMs' embeddings are the next leap: every single word now has a unique identifier—one that captures the word in the context of every other word in every language. He adds: "no, this doesn't automatically lead to all of our deaths, that is such a wild and unfounded jump."

Original post →

More from AGI Musings

AGI Musings channel →