From: mk_thisisit
Biological processors are a groundbreaking innovation, made from living cells [00:00:01]. Final Spark, a Swiss biocomputing startup founded in 2014, is the first in the world to build a living processor from living neurons [00:01:05]. The company aims to combine the computing power of computers with the efficiency of the human brain [00:01:16]. Their unique approach involves using human brain organoids for computation [00:02:26].
What is a Human Brain Organoid?
A human brain organoid is a small fragment of nervous tissue composed of neurons derived from humans [00:01:26]. These organoids are grown on special plates that allow them to form a three-dimensional structure [00:00:10] [00:04:55]. While many people globally are working on human brain organoids, Final Spark’s unique contribution is their application for computing [00:02:08] [00:02:29]. Functionally, these organoids resemble the fetal brain [00:05:27].
Sourcing and Creating Neurons
Human cells are purchased from specialized companies [00:03:04]. These are pluripotent stem cells, typically derived from fibroblasts (skin cells) that are reprogrammed to a state of pluripotency, meaning they can become any type of cell [00:03:10]. Specific methods allow skin cells to be differentiated or restored to pluripotency [00:03:36]. From this stage, they are transformed into neurons by giving them specific molecules in the culture medium that induce them to become neurons [00:03:47]. The cells then reorganize into structures resembling parts of the brain [00:05:19].
Sustaining the Organoids
In the cell culture room, brain organoids are grown in incubators with precise temperature and carbon dioxide concentrations [00:04:34]. A special pinkish culture medium, containing nutrients, is provided to help them survive [00:05:51]. A flow system with tubes and a pump ensures the medium flows under the organoid at approximately 30 to 50 microliters per minute [00:06:23]. A controlled atmosphere maintains enough oxygen and the optimal temperature of 37 degrees Celsius [00:06:45]. Final Spark has managed to keep these cells alive for almost three years [00:07:03].
Connecting to the Neural Platform
Organoids are placed on a multi-electrode array (MEA) to record their activity [00:07:17]. This system has many different electrodes [00:08:02]. A membrane with a hole, called “confetti,” helps position the organoid exactly in the center and reduces system noise [00:08:39].
The Butterfly Experiment
The “butterfly experiment” demonstrates real-time interaction with the brain organoids in the laboratory [00:11:16]. When a blue dot indicates stimulation is sent, a visual illustration of a butterfly moves in three dimensions [00:11:35]. If the butterfly responds to stimulation, it flies straight towards the light; otherwise, it moves randomly [00:12:08]. A response indicates neural activity, likely a “burst” where multiple neurons are transmitting information [00:12:20]. This demonstrates established communication with the organoids [00:12:47].
To induce neurons to process information and enforce behavior (learning), dopamine is used to reward the organoid if it performs a desired action [00:13:03].
Why Biocomputing?
The initial motivation for Final Spark’s project stemmed from an engineering desire to solve problems with the best solutions [00:09:12]. Previous work with artificial neural networks revealed their immense energy consumption, making them difficult to scale [00:09:34]. A simulation of 100 artificial neurons could consume several kilowatts of power, whereas the human brain, with 86 billion neurons, requires only 20 watts [00:10:01]. This stark difference highlights the incredible energy efficiency of biological systems [00:10:07].
The main reason for pursuing biocomputing was the realization that using living neurons was the best way to advance artificial intelligence [00:10:31]. Biological processors are groundbreaking because they are made from living cells, something unprecedented in history [00:10:53].
Challenges and Progress
A significant challenge has been simply keeping the organoids alive; initially, they would die after a few hours [00:14:05]. Progress in biocomputation is slow because experimental results require rigorous confirmation [00:14:49].
A positively surprising moment for the team was successfully recording neuronal activity and growing the cells themselves from liquid nitrogen [00:16:19]. Observing the response of neural tissue to dopamine, showing a wave of activity typical of dopamine-induced stimulation, was a key success [00:16:34].
Ethical Considerations
Final Spark recognizes the ethical implications of using living neurons for computational purposes [00:14:46]. They have established contacts with universities and their ethics experts, even attending ethics conferences to introduce their work to ethicists [00:14:57].
The decision to use human cells over animal cells (like rat cells) for computations is strategic [00:15:32]. While rat cells might work for computation, the potential for therapeutic applications is a significant factor, as discoveries could help people more than rats [00:15:41].
Future Outlook
Biocomputing is seen as the future of data processing and computing, especially for artificial intelligence calculations [00:16:56]. Just as quantum computers are limited to certain computation types, biocomputing can be used for specific calculations, with AI fitting perfectly due to its basis in simulations of neurons [00:17:10]. This project has the potential to transform AI development and drastically reduce its carbon footprint [00:17:29].