Scientists Are Building Artificial Brains From Living Cells

Why It Matters

If biological neural networks can be scaled, they could dramatically cut the power draw of AI while offering insights into brain function, reshaping both computing and neuroscience research.

Key Takeaways

• Polymer mesh scaffold supports tens of thousands of neurons
• 3D‑MIND records action potentials with minimal energy use
• Rat hippocampal neurons form stable 3‑D connections
• Scaling to human‑level networks remains a major challenge

Pulse Analysis

The quest for artificial intelligence that rivals the human brain has long been hampered by the massive energy demands of silicon‑based neural networks. Modern AI models can consume megawatts of power, prompting researchers to explore biologically inspired alternatives. Princeton’s 3D‑MIND platform bridges living neurons with a programmable electronic scaffold, creating a hybrid system that mimics the brain’s three‑dimensional architecture while operating on a fraction of the electricity required by conventional hardware. By embedding electrodes and metal micro‑wires into a flexible polymer mesh, the team achieved reliable growth of rat hippocampal neurons, allowing continuous monitoring of electrical activity and demonstrating that living cells can serve as low‑power computational units.

Beyond energy efficiency, biological neural networks promise a new level of computational versatility. The brain’s dense and sparse connectivity patterns enable rapid, parallel processing and adaptive learning—features that are difficult to replicate with purely digital circuits. The 3D‑MIND device captures these dynamics by preserving the natural 3‑D arrangement of neurons and supporting glial interactions, which are essential for nutrient delivery and signal modulation. This biologically faithful environment not only improves the fidelity of neural signaling but also provides a platform for studying brain physiology and pathology, potentially accelerating drug discovery and neuro‑engineering breakthroughs.

Scaling the technology remains the critical next step. While the current prototype demonstrates feasibility with rat neurons, translating the approach to human‑derived cells and expanding the network to computationally relevant sizes will require advances in scaffold manufacturing, long‑term cell viability, and integration with existing digital infrastructure. If these challenges are met, hybrid bio‑electronic computers could usher in a new era of sustainable AI, offering unprecedented processing power without the carbon footprint of today’s data centers. The convergence of neuroscience and engineering embodied in 3D‑MIND marks a pivotal moment in the evolution of computing technology.