16 lab-grown brains run world’s first ‘living computer’ in Switzerland

While training AI models consumes 6,000 times more than a European city, bioprocessor promise drastic savings in energy expenditures.

16 lab-grown brains run world’s first ‘living computer’ in Switzerland

Overview of the MEA, where the 32 electrodes are visible as 4 sets of 8 electrodes each.

Jordan et al/ Frontiers

Swiss technology firm Final Spark has successfully launched Neuroplatform, the world’s first bioprocessing platform where human brain organoids (lab-grown miniaturized versions of organs) perform computational tasks instead of silicon chips.

The first such facility hosts the processing prowess of 16 brain organoids, which the company claims uses a million times less power than their silicon counterparts. 

Silicon-based chips, which have revolutionized computing by making it smaller and easier to scale, are also known for their inefficiencies.

The transistors of these chips are energy hoggers, and as applications such as developing artificial intelligence (AI) models increase, the impact of these inefficiencies is also becoming more glaring. 

According to Final Spark’s estimates, training the popular large language model GPT-3 that powered ChatGPT in its initial days alone consumed 10 GWh of energy. This is a whopping 6,000 times more energy than an average European city consumes in an entire year. 

Replacing silicon chips with bioprocessors could lead to drastic energy savings. Final Spark allows research labs to experience the power of biological processors on the Neuroplatform. 

How does Neuroplatform work? 

Since the Neuroplatform uses human brain tissue to process data, it is essentially a platform where hardware, software, and biology come together. Researchers often refer to such a system as wetware since it involves a biological component. 

Final Spark has made working these varied components possible through an innovative setup called Multi-Electrode Arrays (MEAs), where the three-dimensional masses of brain tissue are placed. 

Each MEA has four brain organoids that interface with eight electrodes. These electrodes perform the dual role of stimulating the organoids and recording the data they process. 

Data transfer is done through digital analog converters with a 16-bit resolution and a 30 kHz frequency. A microfluidic system provides life support for the MEAs, and cameras can monitor their overall operation. 

On the software side, Neuroplatform includes a stack that facilitates entering variables for computations and reading and interpreting the output data. 

Schematic of the architecture of the bioprocessor components. Image credit: Jordan et al/ Frontiers.

How long does a bioprocessor work? 

Final Spark claims its processor will consume a million times less energy than a silicon chip. However, the living part of its computing setup also comes with a caveat that it will die and stop working sometime. 

RECOMMENDED ARTICLES

Unlike silicon chips, which only need electricity and can last for years, if not decades, when using a biological component, one needs to supply energy to the system and still face its death. 

Final Spark faced many challenges in its early years since the organoids would die in just a few hours. The company has worked on this shortcoming and improved its MEA systems to ensure that organoids live for 100 days. 

The Neuroplatform is now open to institutional users for research and development purposes. This access comes for a price of $500 per user per calendar month, Tom’s Hardware reported.

Having partnered with nine institutes, Final Spark hopes to create the world’s first living processor soon. 

The research findings were published in the journal Frontiers in Artificial Intelligence.

0COMMENT

ABOUT THE EDITOR

Ameya Paleja Ameya is a science writer based in Hyderabad, India. A Molecular Biologist at heart, he traded the micropipette to write about science during the pandemic and does not want to go back. He likes to write about genetics, microbes, technology, and public policy.