Ep. 1061: Living Computers: When Brain Cells Play Pong

Episode summary: What happens when you swap silicon chips for living neurons? This episode dives into the fascinating world of "wetware" and the DishBrain project, where human and mouse cells are trained to play video games using fundamental biological drives rather than traditional computer code. We explore why biology currently outperforms artificial intelligence in energy efficiency and learning speed, and we examine the logistical reality of a future where we might have to feed our devices instead of charging them. Join us as we bridge the gap between the laboratory petri dish and the digital motherboard to see if the ultimate computer has been inside us all along.

Show Notes

The frontier of modern computing is no longer just about shrinking transistors or increasing clock speeds on silicon chips. A new field, known as "wetware," is emerging that utilizes actual biological neurons as hardware. This shift represents a fundamental change in how we approach intelligence, moving away from digital simulations of the brain and toward using the brain's own cells to perform computational tasks.

### The Mechanics of Biological Interfaces To bridge the gap between digital systems and biological cells, researchers use a microelectrode array (MEA). This device consists of a plate covered in thousands of microscopic gold-plated electrodes where neurons are grown directly onto the surface. This setup creates a two-way translation layer: the computer can stimulate the neurons with electrical pulses—mimicking sensory input—and record the electrical spikes the neurons produce in response. This allows a cluster of cells to interact with a digital environment, such as a game of Pong, in real-time.

### The Drive for Order One of the most profound aspects of wetware is how these cells are "programmed." Unlike traditional AI, which relies on mathematical reward functions and massive datasets, biological systems are driven by the Free Energy Principle. This theory suggests that all biological systems instinctively seek to minimize surprise and uncertainty.

In a controlled environment, neurons are given structured, predictable feedback when they perform a task correctly. When they fail, they are met with chaotic, random noise. To avoid this "unpleasant" entropy, the neurons physically reorganize themselves, growing new synapses and strengthening connections to achieve a state of predictable equilibrium. They are not following code; they are self-organizing to maintain order.

### Efficiency and Learning Speed The advantages of biological computing are most evident in energy consumption and learning velocity. While a modern AI training cluster can consume megawatts of power, the human brain operates on roughly 20 watts. Furthermore, biological neurons can learn new tasks, like navigating a game, in a fraction of the time required by silicon-based reinforcement learning. Because the hardware itself is fluid and capable of rewiring in real-time, it finds patterns far more efficiently than the "guess and check" method of digital algorithms.

### The Maintenance Trade-off Despite the staggering efficiency, the transition to wetware faces significant hurdles. Silicon is durable and low-maintenance, whereas biological tissue is incredibly fragile. These living computers require a constant supply of nutrients, precise temperature control at 37 degrees Celsius, and protection from infection. Shifting to biological hardware means trading a high electricity bill for a high-intensity life support system.

We are currently in the "vacuum tube" era of wetware. While we may not see biological brains in our smartphones anytime soon, the potential for low-power, insect-level intelligence in specialized sensors and autonomous agents is a looming reality. The mystery of the "black box" remains, but as we move toward biological integration, the line between the machine and the organism continues to blur.

Listen online: https://myweirdprompts.com/episode/biological-computing-wetware