Scientists have discovered a way to integrate living human brain cells into computing systems, which may make the “A” in AI obsolete.
Research paper officially published today in the journal natural electronicsResearch by scientists at Indiana University Bloomington has described a new system called “Brainoware” that uses the human brain organism to complete advanced AI tasks. These organisms, which are artificially grown clumps of cells or tissue that resemble organs, are now mounted on high-density multi-electrode arrays, which are quite primitive today. But researchers hope that using this technology could pave the way for biocomputers that can perform the same tasks that computers do, but with minimal energy consumption.
“The human brain typically consumes about 20 watts, while current AI hardware consumes about 8 million watts to power a comparative artificial neural network (ANN),” the research paper claims. “Brainoware can provide additional insights into AI computing because brain organisms can provide biological neural networks (BNNs) with complexity, connectivity, neural plasticity and neurogenesis, as well as low energy consumption and fast learning capabilities.”
About “Brainoware”: A miniature brain-like structure made from human cells, known as a brain organism, used as a living AI to perform tasks such as solving complex equations. 🧵 1/8https://t.co/nuCNOWNf3j
— Michael Le Page (@mjflepage.bsky.social) (@mjflepage) March 14, 2023
“Because the human brain uses much less energy and learns much faster, some researchers see biocomputing as the way forward,” Michael Le Page wrote on Twitter in March, but the field has its limitations. He pointed out that if pushed this far, difficult questions could be raised.
Le Page quotes Cambridge developmental neurobiologist Madeline Lancaster. “For these issues to go beyond ethical boundaries is something we definitely want to avoid, and the scientific and ethical communities are coming together to define where those boundaries are.”
Brainoware sends and receives information from the brain organism through “adaptive storage computation.” This method enables unsupervised learning from training data, which can still shape the functional connectivity of the organoids. The practical potential of the system has been demonstrated in tasks such as speech recognition, where after training it achieved increased accuracy in distinguishing the voices of individual speakers.
For example, organoids were trained to identify a single individual’s voice from a set of 240 audio clips of eight people pronouncing Japanese vowel sounds. After training, the organoids were able to complete the task with over 70% accuracy.
But science still has a long way to go in creating living robots. Organoids can only identify speakers and cannot understand speech. That means we have a very long and winding road before this technology can actually be used in medicine or engineering.
Titouan Parcollet from the University of Cambridge said: new scientist The magazine said the potential of biocomputing was enormous, but acknowledged that “current deep learning models are actually far better than any brain at certain targeted tasks.”
The researchers also cautioned that “current organoids still suffer from high heterogeneity, low production throughput, necrosis/hypoxia, and variable viability,” making them unfeasible for use other than current research purposes.
Parallel to Brainoware’s development, AI has been creatively applied to fields such as healthcare, with innovations that help restore mobility to quadriplegics and AI models that can read minds. Together, these developments highlight the versatile and innovative nature of AI technology.
Edited by Ryan Ozawa.
