The Robots Are Already Here—The Data Gap Is What’s Holding Them Back

About this episode

What happens when robots stop looking like industrial machines—and start looking (and even feeling) human? And if “replicants” become plausible within our lifetimes, what would it take to get there… and what might it break along the way?

In this episode of The People’s AI, presented by the Vana Foundation, we explore the robot revolution from three angles: what robots can actually do today (quietly, at scale), what’s likely in the near-term (especially in warehouses, logistics, healthcare, and elder care), and what the more radical futures imply—humanoids, “fleshbots,” and the thorny question of rights and personhood. 

A through-line across every conversation: the hidden constraint isn’t just hardware or dexterity—it’s data. Robotics doesn’t have an LLM-sized training corpus, and that gap shapes everything from progress timelines to privacy concerns and labor dynamics. We also dig into an under-discussed limiter: power consumption, and why energy efficiency may quietly govern how ubiquitous robots can become.

Guests

• Thomas Frey — Futurist (former IBM engineer)
• Dr. Aniket Bera — Director of the IDEAS Lab at Purdue University
• Jeff Mahler — Co-founder & CTO, Ambi Robotics

What we cover

• Why most impactful robots won’t look humanoid (at least at first)
• Specialized machines—crane-like systems, warehouse sorters, mobile carts—are already delivering value because they can be engineered for reliability in constrained environments.
• The robots already among us (even if we don’t notice them)
• Warehousing and supply chain, recycling and waste sorting, mobile delivery systems, and surgical robotics are all expanding—often out of public view.
• Humanoid robots: where they might actually make sense
• Homes, hospitals, assisted living, and caregiving settings—places where human spaces and human expectations matter—may be the earliest “real” markets.
• Robots in science and medicine: the bullish case
• Lab automation, drug discovery loops, high-throughput testing, and more precise (and potentially remote) surgical procedures could be some of the most meaningful gains.
• The true bottleneck: the robot data gap
• LLMs feast on web-scale text. Robots need massive volumes of real-world interaction data—vision, touch, force, motion, and the consequences of actions.
• How robot companies may collect data (and what that implies)
• Motion-capture / imitation learning (wearables that mirror human movement), teleoperation (“humans in the loop” controlling robots remotely), simulation, and deployment flywheels that generate production data.
• Privacy + labor: the coming debate
• If robots learn from human environments and human demonstrations, who owns that data—and who gets paid for producing it?
• A final irony: why humanoids might win more share than we expect
• We have endless data of humans doing tasks—videos, demonstrations, routines—so humanoid form factors may benefit from transfer learning advantages, even if they’re not mechanically optimal.

About Vana

The People’s AI is presented by the Vana Foundation, supporting a new internet rooted in data sovereignty and user ownership—where individuals, not corporations, govern their own data and share the value it creates.

Learn more at Vana.org.