They create a biological computer by connecting 16 human minibrainsCredit: Jordan et al., Frontiers in Artificial Intelligence, 2024.
A Swiss startup has just launched a biological computer that connects to living, pulsating brain cells: it uses much less energy than traditional bit-based computers, in a new quest to achieve computing devices and Artificial Intelligence (AI) applications that can achieve a functioning efficiency similar to that obtained by the human brain.
Related
The Swiss firm FinalSpark has created an online platform called Neuroplatform, which uses spherical clusters of human brain cells grown in the laboratory, known as brain organoids or minibrains. A total of 16 organoids are housed in four arrays, which connect to eight electrodes each and a microfluidic system intended to deliver water and nutrients to the cells.
More powerful, efficient and sustainable
According to an article published in Impact Lab, bioprocessors like those used in Neuroplatform consume a million times less power than conventional digital processors, which promises a significant reduction in the environmental impact of computing. At the same time, it is an advance on the path to achieving computing devices with an efficiency similar to that achieved by the human brain.
For decades, computer specialists have sought to emulate the conditions of our brain, replicating its neural networks to create Artificial Intelligence (AI) applications with greater processing power. However, although these artificial neural networks have advanced significantly and have become very powerful, its energy consumption has also increased.
Considering our society’s increasing dependence on these computing and AI devices, their use has become unsustainable in the long term. According to the researchers behind the new biological computer, who summarized their findings in a study published in the journal Frontiers in Artificial Intelligence, training a single large language model (LLM) like GPT-3 required approximately 10 gigawatts/hour, around 6,000 times the annual energy consumption of an average European citizen.
Hardware, software and biology
Scientists maintain that the use of bioprocessors could significantly reduce this energy expenditure, opening a new era in the development of computing devices and AI solutions with greater processing capacity, efficiency and sustainability. Compared to the consumption of GPT-3, the human brain operates 86 billion neurons using only 0.3 kilowatts/hour, as indicated in an article published in Science Alert.
Neuroplatform works through an architecture classified as “wetware”, that integrates hardware, software and biology. The main innovation of the platform lies in the use of four multi-electrode arrays (MEAs), which house living tissue organoids. These are three-dimensional cellular masses of brain tissue. Each MEA contains four organoids: data exchange is facilitated by analog-digital converters. Additionally, software allows researchers to enter data variables and interpret the processor’s output.
Reference
Open and remotely accessible Neuroplatform for research in wetware computing. Fred D. Jordan et al. Frontiers in Artificial Intelligence (2024). DOI:https://doi.org/10.3389/frai.2024.1376042
