According to its own statement, Swiss startup FinalSpark aims to be “Bioprocessor“I made it out of 16 Organic to the human brain It is composed. Embedded bioprocessors like these are said to “use a million times less power than traditional digital processors,” according to the company.
➤ Read more: Dream of a brain in a jar
Organoids are made from a population of stem cells and are designed to mimic the structure and function of an organ such as the brain. high FinalSpark is their neural platform capable of obtaining information For learning and processing – And with high efficiency.
Electrons stimulate cells and record signals
The company's main innovation consists of 4 multiple electrodes containing living tissue. Each of these multiple electrodes contains 4 organelles, each connected to 8 individual electrodes. These electrodes can affect both brain cells Stimulating and recording electrical signals from it.
➤ Read more: One million implanted brain cells learn to play ping-pong
Organisms exist in a nutrient solution that keeps them alive. However, while digital computer chips last for years and decades, bioprocessors can only last for approx. 3 months Uses.
It's still at the beginning of the research
“Imagine the state of research in quantum computing 20 years ago — this is where we are today with biocomputing,” describes co-CEO Fred Jordan in a podcast. There are currently only 3 companies in the world that deal with bioprocessors.
In the future, bioprocessors will face the increasing power demand of IT systems. Just training a language model like GPT-3 takes it 10 GWh This is approximately 6,000 times the energy consumption of the average European citizen. With bioprocessors, this would be an expenditure of energy Slim.
Artificial brains need training
However, there is still a long way to go until then. Like AI models, neurons in organisms have to do this as well Become an apprenticeTo provide useful outputs. To date, FinalSpark is working with several universities to develop research in this area.
Jordan finds it interesting to make attempts to recreate biological processes in our brain using classical artificial intelligence – keyword “Neural networks“However, we are now trying again to transfer computer operations to biological cells.
But at the same time, there are also efforts Adapting devices further to biological neurons. The so-called neurochips follow the same approach as the FinalSparks bioprocessors, but without the biological component. They must also be trained depending on the application, and in the best case, they promise huge energy savings and much faster computation times.
➤ Read more: The Austrians are working on computer chips that work like a brain
