The line between science fiction and reality has just blurred significantly. FinalSpark, a technology startup based in Switzerland, has launched an online platform that allows researchers to remotely access a processor made of living human brain cells. This âneuroplatformâ represents a major leap in biocomputing, aiming to solve the massive energy consumption crisis currently facing the artificial intelligence industry.
For decades, computing has relied on silicon chips. These chips are incredibly fast but also incredibly inefficient compared to biological systems. FinalSpark is pioneering a different approach known as âwetware.â This involves combining hardware, software, and biology to create processors that function more like a human mind and less like a calculator.
The centerpiece of this new technology is the Neuroplatform. It is the worldâs first online platform that provides access to biological neurons in vitro. The system uses 16 human brain organoids. These are essentially tiny, lab-grown masses of cells that mimic certain structures and functions of the brain. They are not full brains and do not possess consciousness, but they do exhibit electrical activity and the ability to form new connections.
The physical setup of the FinalSpark processor is a marvel of bio-engineering. It does not look like a standard server rack. Instead, it relies on a sophisticated life-support system.
The most significant breakthrough FinalSpark has achieved is longevity. In the past, experiments with brain organoids typically lasted only a few hours before the cells died. FinalSpark has developed a closed-loop system that keeps these mini-brains alive and functional for up to 100 days. This extended lifespan allows researchers to run experiments that span weeks rather than hours.
The primary motivation behind this technology is energy efficiency. As Artificial Intelligence models like ChatGPT (OpenAI) and Gemini (Google) become more powerful, their power requirements skyrocket.
Training a single large language model (LLM) like GPT-3 requires roughly 10 gigawatt-hours of electricity. That is about 6,000 times the energy a European citizen uses in an entire year. By comparison, the human brain performs complex processing tasks, drives a car, learns new languages, and manages bodily functions all while running on approximately 20 watts of power. That is barely enough to power a dim lightbulb.
FinalSpark estimates that their bioprocessors could essentially consume one million times less energy than traditional digital processors for similar learning tasks. While silicon chips process digital code (0s and 1s), biological neurons process information through chemical and electrical signals across synapses. This biological approach is naturally optimized for efficiency.
You cannot simply install Windows or Linux on a cluster of brain cells. Programming the Neuroplatform requires a completely different paradigm. Instead of writing code, researchers âtrainâ the organoids using operant conditioning, similar to how you might train a dog.
The process involves stimulation and feedback:
Over time, the organoid physically changes. Synaptic connections strengthen or weaken based on this feedback. This is neuroplasticity in action. The biological network reconfigures itself to perform the task more efficiently to receive the reward.
FinalSpark is not keeping this technology behind closed doors. They have opened the Neuroplatform to research institutions globally. The goal is to create a collaborative environment where scientists can develop the first protocols for biocomputing.
Current access is priced for academic budgets. Reports indicate that research groups can subscribe to the platform for approximately $500 per month. This fee grants them 24â7 remote access to the organoids, allowing them to run Python scripts that interact with the biological hardware via an API.
While the potential is enormous, wetware computing faces significant hurdles.
The launch of the FinalSpark Neuroplatform marks a specific moment in history where biology becomes a legitimate component of our technological infrastructure. It offers a potential path away from the unsustainable energy demands of modern AI, trading raw speed for extreme efficiency.
Are these organoids conscious? No. The organoids used by FinalSpark are roughly 0.5 millimeters in diameter. They lack the structural complexity, sensory organs, and neural volume required for consciousness, feelings, or pain. They are essentially tissue cultures that exhibit electrical reactions.
Why use human cells instead of mouse cells? Human neurons are generally more efficient and robust than rodent neurons for these specific types of connections. Using human tissue also provides better data for researchers looking into medical applications, such as understanding how drugs affect neural networks.
Can I buy a biological computer for my home? Not anytime soon. The maintenance required for these processors is intense. They need a constant supply of nutrient fluids, strict temperature controls, and sterile environments. For the foreseeable future, this technology will remain in cloud-based laboratories accessed remotely.
What happens when the organoids die? When the organoids reach the end of their lifespan (currently around 100 days), the system automatically replaces them with new ones grown in the lab. This ensures that the computing power remains available for researchers without long interruptions.