We keep talking about smarter models and better benchmarks. But one big question that we don’t see many talk about is whether AI actually understands the world it’s being asked to operate in.

In this episode of Machine Dreams Podcast, we sit down with Ofir Mulla, CTO of AI video security leader Lumana, to unpack why visual AI still misreads everyday reality and what it would take for machines to truly understand context, behavior, and intention.