Michael Levin: Bioelectric Signaling, Agential Materials, and Robot Scientists

The A-Life Mindset and Embodied Intelligence

• 💡 The A-life mindset is uniquely positioned to advance humanity by understanding how embodied minds exist in the physical universe.
• 🧠 The speaker's group uses AI and computational tools to communicate with biological systems across all scales, from molecular networks to entire organisms.
• 🎯 A key goal is to develop tools for detecting, communicating with, and ethically interacting with a wide range of diverse intelligences.

Agential Materials and the Anatomical Compiler

• 🔬 Agential materials refer to living matter that offers autonomy and intrinsic motivation, requiring communication rather than micromanagement from engineers.
• 🚀 The concept of an anatomical compiler envisions a system where one can design any organism, and it translates that design into stimuli for cells to build it, crucial for applications like regenerative medicine.
• 🔑 This compiler acts as a translator of goals between the engineer and the goal-directed living system, rather than a micromanaging 3D printer.

Biological Plasticity and Multiscale Problem Solving

• 🌱 Biology exhibits a multiscale competency architecture, with problem-solving capabilities and intelligence embedded at every level, including molecular networks and individual cells.
• ✨ Living matter demonstrates extreme plasticity, such as tadpoles seeing with an eye on their tail or planaria retaining memories after brain regeneration, showcasing zero-shot learning.
• 🔄 Memories in biology are not static but transform and remap to new body architectures, as seen in caterpillars becoming butterflies, adapting to new contexts and embodiments.

Exploring Latent Morphological Space

• 🧬 The speaker's work explores the latent space of possibilities for biological forms, beyond what evolution has traditionally produced, using novel systems like Xenobots and Anthrobots.
• 🤖 Xenobots, self-assembling and self-motile biological robots from frog cells, can even perform kinematic self-replication from loose materials.
• 🩹 Anthrobots, derived from adult human tracheal cells, self-assemble into creatures that can heal neural wounds in vitro, demonstrating unexpected capabilities and significant gene expression changes.

The Mambot: An Embodied AI Scientist

• 🧪 The Mambot is an embodied AI robot scientist designed to make hypotheses about cellular stimuli, test them, and refine its understanding of the morphogenetic code.
• 🛠️ This system embodies the active inference cycle of science and agency, continuously making predictions, observing outcomes, and revising its internal models.
• 🌐 Mambot represents a communication interface between human scientists and cellular collective intelligence, aiming to foster an ethical symbiogenesis with novel intelligences.

Future of Human-AI-Biological Collaboration

• 🎯 Future applications include developing new synthetic living machines for repair, cleanup, and exploration, and using robot science to crack the morphogenetic code for regenerative medicine.
• 🤝 A critical long-term goal is to enable the Mambot to develop a curiosity loop to pose its own problems and even communicate with bots to understand their 'wants' and goals.
• 🚀 This research aims to develop tools for recognizing and relating to radically different kinds of minds, fostering collaborative and aligned systems for a mature society.