What happens when you stop selling keywords and start selling embeddings; the mathematical fingerprints of who a user is? According to Gilad Lotan, Head of AI/Data Science & Analytics at BuzzFeed, entire industries from adtech to media could be turned upside down. In this episode of Almost Human, Eden Shochat continues his conversation with Gilad to unpack why embeddings are one of the most under-hyped yet world-shaping ideas in AI.

The two dive into why embeddings act like machine code for models - unlocking new capabilities that are both powerful and unsettling. Along the way, they explore the geopolitical race between the U.S. and China, the battle between Google and OpenAI, and how hyper-personalized ads might redefine digital business models. Gilad shares what tools BuzzFeed’s AI team is really using, like “Replicate” for rapid model testing, and what’s still missing from the AI ecosystem, from true fine-tuning to rigorous evaluation frameworks.

To close things out, Eden challenges Gilad to a rapid-fire round of questions: When will AGI arrive? Who’s really ahead? What’s the next winning ad unit? The answers may surprise you.

Whether you’re an operator, researcher, or investor, this episode offers a sharp, candid look at how embeddings (and the infrastructure around them) are quietly reshaping the future.

Please rate this episode 5 stars wherever you stream your podcasts!