Morphospace / Life, Machines, and a Foundation for Synthbiosis

Life, Machines, and a Foundation for Synthbiosis

In one paragraph

A compressed tour of the lab's program: goals encoded bioelectrically, memory in xenobots, anthrobots from human cells, learning in tiny gene-network models, and the Mambot robot scientist.

Main ideas (plain words)

  • The future is not humans versus AIs: almost any mix of evolved material, engineered material and software can be a viable being.
  • Against the standard view that cells 'know nothing', Levin says set points are real because the lab can read and write them bioelectrically.
  • Xenobot signaling data differ from null models about as much as human fMRI data do; unpublished work suggests xenobots can hold at least two memories for at least 24 hours.
  • Small gene-regulatory-network models (a handful of equations) can learn — a 'free gift from math', not from physics or evolution. Even simple sorting algorithms show unexpected competencies.
  • Mambot: 'somewhere between a tool and a colleague' — a robot that builds xenobots, observes them, and revises its hypothesis, i.e., the full loop of science in morphospace. He closes with three disclosures of companies that have licensed the lab's technology.

Why it matters

Best short introduction to send someone new. Mambot is an AI system running the full experiment loop.

Key quotes (word for word from YouTube's auto-generated captions)

Quotes are copied exactly from the captions, including filler words and caption errors. Timestamps are to the nearest minute.

"this is not about humans versus AIs or language models or anything like that"
— 0 min

"We've developed some of the first tools to read and write pattern memories in tissues outside of the brain."
— 3 min

"This is somewhere between a tool and a colleague."
— 18 min

Related

What's new · Companies tracker · transcript: raw/transcripts/2026-09-22-6W5qhMasROY.txt