Why now? After decades of false starts, robotics is finally entering its exponential phase and the reasons go far beyond better models.
In this insightful fireside chat, OpenAI’s Caitlin Kalinowski (formerly Head of Hardware at Meta and Apple) joins Maryanna Saenko (Co-founder of Future Ventures) to explore why spatial AI, reinforcement learning, and next-gen hardware are colliding at the right moment.
They discuss:
• The hardware renaissance enabling general-purpose robotics
• Why sim-to-real ratios are changing everything
• RL’s power to reduce component costs (and raise capabilities)
• What COVID taught us about global manufacturing resilience
• The future of robotics supply chains beyond China
• How AI is 4x-ing learning velocity for hardware teams
Timecodes:
00:07 – Intro: from motorcycling buddies to AI hardware leaders
01:00 – Caitlin’s journey: Apple, Meta, Quest, OpenAI
02:48 – Why robotics is ready now: convergence of 3 acceleration curves
03:33 – RL is replacing classical control systems
04:22 – Sim-to-real is finally tipping the balance
05:51 – Simulated data vs. PhDs folding T-shirts
06:35 – How smarter software changes hardware design strategy
07:28 – Future-proofing robotics hardware for tomorrow’s AI
08:54 – Will cheap AI compensate for bad hardware? Not quite
09:45 – Why supply chains (not AI) keep Caitlin up at night
10:28 – Beyond China: hybridization and tariff-era design
12:02 – How Meta shipped Quest during COVID disruptions
13:30 – Automated factories are already here
14:36 – Why the US struggles with flexible manufacturing
15:25 – Robotics still constrained by physical space
16:14 – Has AI changed hiring? Not yet — but it’s 4x-ing learning speed
16:58 – Software teams will see more AI impact than hardware (for now)
17:39 – You want agents, but more than that — you want taste18:30 – Wrap-up: excitement, friendship, and quiet expectations
