AI’s Next Phase: From Research to Real-World Value

AI’s Next Phase: From Research to Real-World Value

May 8, 2025 · 25 min

People

Watch on

Suggest an edit

AI is no longer just evolving — it’s accelerating into production. At Madrona’s 2025 Annual Meeting, Jon Turow sat down with AI2 CEO Ali Farhadi, Meta/UW’s Luke Zettlemoyer, and Brian Raymond, CEO of Unstructured, for an unfiltered look at AI’s next phase. What’s changing: Research is moving from paper to product in a matter of hours Multi-agent systems are no longer vaporware — they’re driving real companies The best AI reasoning is happening in narrow, deeply tuned domains Open source isn’t just philosophy — it’s a requirement for tuning and trust This is a must-watch if you’re building, investing, or deploying AI infrastructure. Learn how the fusion of open research, developer energy, and enterprise-grade tooling is reshaping what’s possible — and what’s next. Subscribe for more conversations from the front lines of AI innovation. Chapters: 00:00 — Introduction 01:00 — Research to Production in 24 Hours 03:00 — Rise of Multimodal and Multi-Agent Systems 06:00 — State of Reasoning in AI Today 11:00 — Vertical AI vs. Generalized AI: What’s Working 16:00 — Breakthroughs in Vision-Language Models 19:00 — Why Open Source Still Matters 24:00 — The Next Phase: Customization, Tuning & Real-World Value

AI2 Meta Madrona 2025 Annual Meeting Unstructured Madrona

Chapters

  1. 0:00 Introduction
  2. 1:00 Research to Production in 24 Hours
  3. 3:00 Rise of Multimodal and Multi-Agent Systems
  4. 6:00 State of Reasoning in AI Today
  5. 11:00 Vertical AI vs. Generalized AI: What’s Working
  6. 16:00 Breakthroughs in Vision-Language Models
  7. 19:00 Why Open Source Still Matters
  8. 24:00 The Next Phase: Customization, Tuning & Real-World Value