RoboBoom
The Humanoid Hub· November 18, 2026· 37:16

Why Robots Shouldn’t Copy Humans | Jan Liphardt

Jan Liphardt, founder of OpenMind, argues that physical AI should be designed around specific jobs rather than copying human bodies, speaking at a Humanoid Hub event.

In this talk

  1. 0:41

    Design for the customer’s task

    Liphardt says customers usually start with a budget, a task, and reliability requirements, then ask for a lower price. That pressure leads teams to remove features that are not essential. He sees the opportunity in choosing physical forms that do a job more safely, quickly, and cheaply than a human-shaped robot.

  2. 3:40

    Expect many specialized robot forms

    Using AlphaGo’s move 37 as an example, Liphardt asks what physical AI might do that people would not think of themselves. He points to Ukraine’s mix of low-cost, modular drones adapted to specific missions, rather than a Terminator-style humanoid. He expects a wider range of forms, including wheeled robots, tracked machines, and task-specific grippers.

  3. 12:45

    Modularity can improve capability

    Liphardt describes hands with two opposed thumbs, continuous wrist rotation, and quick-change couplings for swapping tools. He also highlights a Swiss robotic hand that can detach, crawl to retrieve objects, and return to the arm. Robots can share accessories, maps, and information across a fleet, enabling forms of cooperation that human bodies cannot offer.

  4. 11:10

    Build around real constraints and users

    Physics and human spaces still matter: robots must account for motion limits, door widths, counters, and handles. Liphardt says humanoids can be a useful engineering goal that drives progress in actuators, batteries, chips, and sensors, while doubting that most future machines will look exactly like people. He urges builders to listen to customers and imagine forms suited to their needs.

  5. 18:09

    Make safety and trust part of deployment

    Liphardt points to Waymo as an example of a quiet, highly safe autonomous system, and predicts that unobtrusive reliability may be a major physical AI advantage. He says homes pose deployment challenges such as inconsistent Wi-Fi, varied doors, and insurance. OpenMind’s open-source hardware abstraction layer is intended to help people inspect and adapt robot software across platforms.

Briefs summarise talks and interviews; timestamps open the video at that moment. Claims in a talk are the speaker's own.

More talks