With 86 billion neurons and a power consumption of only about 20 watts, the human brain is the unrivaled winner when it comes to efficient computing. If we were to attempt to create something similar using silicon, at least 10 megawatts of power would be needed (according to Kwabena Boahen, a computer scientist at Stanford University). So why not seek inspiration from nature and learn to use and control biological neural networks? This is exactly the objective of FinalSpark.
With the discovery of induced pluripotent stem cells (IPSC) by Shinya Yamanaka in 2006, the ethical boundaries of using primary neurons or stem cells were removed, clearing the way for biocomputing using biological neurons. Using IPSCs, it is now possible to grow neurons in cell cultures and use them as computing power.
This sounds easy, but the stumbling blocks along the way are tremendous. Going from a single neuron to a large biological network with billions of neurons is not as easy as mixing a cocktail.
The main challenges are as follows:
- Programming of biological networks: For humans, the process of absorbing information is called learning. In IT, this is called programming. To efficiently interact with and teach neural networks, we are examining novel programming approaches that allow for the targeted and repetitive programming of biological structures.
- Growing large biological networks and keeping them alive: Culturing the right mix of neurons and enabling them to grow into large, vascularized structures are challenging tasks. Today, only small volumes of up to 1 mm3 can be successfully produced and kept alive. Fundamental R&D is necessary to further increase the volume and extend the lifetime.
In our bio lab at FinalSpark, we are actively working on all the aspects mentioned above. Will we succeed? Nature has shown that biological neurons are well suited for intelligent life. Therefore, we are convinced that replicating what nature has done—and then going beyond what we know today—is above all a question of time and money. Silicon-based artificial intelligence is very popular today, but it faces numerous problems when it comes to generalized artificial intelligence and scalability. FinalSpark’s approach using biological neurons avoids these problems and opens limitless new perspectives for enhancing life on earth.