State of the Art of Biological Computing • Ewelina Kurtys & Charles Humble • GOTO 2026
Dr. Ewelina Kurtys of FinalSpark discusses the ambitious goal of building computers from living neurons, aiming for a 1-million-fold increase in energy efficiency over digital systems. The conversation delves into the technical challenges of neural encoding and plasticity, the use of brain organoids, and the profound ethical and philosophical questions surrounding determinism and consciousness in biological hardware.
Dr. Ewelina Kurtys, Strategy Advisor at FinalSpark, outlines an ambitious vision to revolutionize computing by using living neurons as the fundamental hardware. This approach is not an incremental improvement but a paradigm shift, promising a 1 million-fold increase in energy efficiency compared to current digital computers.
The Core Mission: Tackling AI’s Energy Crisis
The primary driver behind biological computing is to address the unsustainable energy consumption and cost of modern AI, particularly Large Language Models (LLMs). While companies currently subsidize the cost of services like ChatGPT to drive adoption, the real operational expenses are enormous and growing with each new, more powerful model.
FinalSpark’s goal is to create a new type of hardware that makes AI drastically cheaper for the end-user. The vision is a general-purpose biocomputer, developed over an estimated 10-year timeline, that could eventually be housed in massive “bioservers.” The project is currently self-funded, but the team is seeking 50 million Swiss Francs to expand the research team and accelerate development.
Performance and Capabilities of Biological Computers
The performance characteristics of a biological computer are expected to differ significantly from silicon-based systems.
- Efficiency over Speed: Biological processors will likely be slower than digital ones but exceptionally efficient at solving complex problems, much like the human brain. They are particularly well-suited for tasks like generative AI.
- Parallel Processing: The brain’s inherent parallelism is a key reason for its energy efficiency, a feature that biological computers would naturally inherit.
- Memory: Similar to the human brain, these systems would excel at conceptual and creative tasks rather than rote memorization of vast datasets.
Technical Challenges and Breakthroughs
Building a functional biocomputer involves overcoming several fundamental scientific and engineering hurdles.
Neural Encoding: The Biggest Hurdle
The most significant challenge is understanding how neurons encode information. Unlike the binary 0s and 1s of digital systems, the brain uses a complex system based on the timing and spatial location of neural activity. We can detect these signals, but we cannot yet fully decode their meaning. This is the core reason why “mind-reading” remains in the realm of science fiction and is a central problem to be solved.
Plasticity and Reproducibility
Neural systems exhibit plasticity, meaning they change and adapt over time. While this is the basis for learning, it presents a major challenge for scientific experiments, as results can vary from one day to the next. To ensure validity, FinalSpark’s experiments require a high degree of automation and must be repeated many times to be considered reproducible.
A notable success has been the ability to reproducibly store one bit of information. This was achieved by mathematically calculating a “center of activity” based on signals from eight electrodes and demonstrating the ability to consistently shift this center between two distinct states.
The Neuroplatform and Organoids
FinalSpark’s current prototypes are built around organoids—3D structures containing around 10,000 neurons each. These organoids are derived from human skin cells that are reprogrammed into stem cells and then differentiated into neurons.
- It’s crucial to understand that these are not “brains on a dish.” They are complex cellular structures that serve as the building blocks for research, but they are vastly simpler than an actual brain.
- The organoids are placed on electrodes, allowing researchers to send electrical signals and measure responses. The system also supports chemical stimulation, such as releasing dopamine to act as a reward signal during experiments.
- This entire platform is accessible remotely, allowing universities and industry clients to conduct fundamental research on biocomputing.
Philosophical and Ethical Dimensions
The work naturally raises profound questions about the nature of the systems being built.
Determinism vs. Free Will
Dr. Kurtys, referencing the work of Robert Sapolsky, argues that the brain’s operations are fundamentally deterministic. While its immense complexity gives the appearance of non-determinism, every action is ultimately the result of a causal chain rooted in genetics and environment. This belief underpins the conviction that a biological computer, while complex, is ultimately a controllable system.
Consciousness and Suffering
The question of whether these neural systems could be conscious or capable of suffering is a primary ethical concern.
- Dr. Kurtys makes a clear distinction between biological stress (any deviation from homeostasis) and suffering, which is viewed as a more complex phenomenon requiring a brain.
- There is currently no scientific way to measure consciousness, and FinalSpark actively collaborates with philosophers to navigate these complex ethical questions responsibly.
- Informed consent for the cells is handled by the commercial companies that supply the stem cells, which are derived from skin samples in a fully regulated process.