Research thread · Physical AI × biology

Bionics
Physical AI.

What happens when Physical AI stops treating the body as a passive container for intelligence — and starts borrowing the logic, materials and interfaces of life?

“Bionics Physical AI” is my working label, not an established discipline. The research behind it crosses soft robotics, morphological computation, neuroprosthetics, biohybrid robotics, developmental biology and adaptive machines.

IMITATEEMBODYINTERFACEINTEGRATEHYBRIDIZEGROW
Where the rabbit hole went next

First the system.
Then the body.

The Robot Is Not the System started from the architecture around intelligent machines: perception, simulation, edge intelligence, orchestration, safety and human authority.

This research follows one thread inward. What if some of the capability is not in the model at all? What if it sits in geometry, compliance, material properties, living tissue, peripheral sensing or a nervous system connected directly to the machine?

Physical AI gave intelligence a body. Bionics asks what that body should be made of — and how much intelligence the body can carry itself.
Start with The Robot Is Not the System
Five-part research series · complete

The body becomes part of the research.

Five designed articles following the body from biological inspiration and morphological intelligence through living machines, neural interfaces and bodies that keep changing after deployment.

Cover of The Body Is Part of the Algorithm
02 / 05
THE BODY AS INTELLIGENCE

The Body Is Part of the Algorithm

This article looks at the intelligence carried by geometry, compliance, materials and local physical dynamics. A body can simplify control, absorb uncertainty, distribute sensing and sometimes participate directly in computation.

Read the articlePDF ↓
Cover of Where the Machine Ends
03 / 05
THE LIVING MACHINE

Where the Machine Ends

Living muscle, neurons, cyborg organisms, Xenobots and Anthrobots pull robotics into a different engineering regime. Components need metabolic support, develop over time and bring biological dynamics into systems that engineers usually try to specify completely.

Read the articlePDF ↓
Cover of When the Robot Becomes Part of You
04 / 05
THE HUMAN-MACHINE LOOP

When the Robot Becomes Part of You

This article follows the boundary from the human side: osseointegrated prostheses, neural control, artificial sensation, embodiment, AI-assisted action and augmentation. The defining architecture is a loop between nervous system, machine and environment.

Read the articlePDF ↓
Cover of After the Robot
05 / 05
THE UNFINISHED BODY

After the Robot

The final article looks at machines whose bodies are no longer fixed: robots that model themselves, adapt to unexpected physics, reconfigure, detect damage, repair and combine different timescales of change.

Read the articlePDF ↓
A useful way to map the territory

Two axes, not one ladder.

Biological embodiment and artificial intelligence do not increase together. A machine can be computationally sophisticated and almost entirely synthetic, or biologically sophisticated with very little conventional AI. The interesting systems begin to spread across both dimensions.

HOW BIOLOGICAL IS THE BODY?
SyntheticBio-inspiredBiologically integratedBiohybridLiving
WHERE DOES THE INTELLIGENCE RESIDE?
Central softwareDistributed controlMorphology + materialBiological sensingNeural / cellular networks

Conceptual placement, not a maturity score. The point is that “more biological” and “more AI” are different dimensions.

What I’m watching

This research is still fragmented across materials, scales, laboratories and applications. The near-term signals are not all moving at the same speed.

  • Already moving on real hardware: shared human–machine control, adaptive control, automated design loops and morphology-aware learning.
  • Earlier: reconfigurable bodies, distributed physical intelligence and autonomous repair.
  • Laboratory frontier: living, self-maintaining machines whose biological components can sustain, remodel and perhaps repair themselves.

The governance questions arrive early: agency in shared-control prosthetics, neural-data ownership, animal welfare in cyborg systems, and what obligations might eventually attach to living components.

The robot may never be completely finished.

Back to Physical AI research