A research team from Indiana University Bloomington has reported the successful creation of a brain organoid-based artificial intelligence processing unit, known as Bainoware. These brain organoids were placed on an electrode array and connected to electrical currents. Data input was then converted into electrical stimuli that were sent to the brain cells, with the resulting response from the cells being read.
The team tested the learning capabilities of these brain cells by training them to predict a non-linear system called the Hénon map using 200 sets of input data. The training process took one day per round, and after four rounds, the regression score for data prediction increased from 0.356 to 0.812. This demonstrates the learning ability of the brain organoid model compared to an artificial neural network model with long short-term memory (LSTM), which achieved a similar score but required 50 rounds of training.
Brainoware showcases the potential of using brain cells to reduce both the time and energy required for training artificial intelligence. In the future, this technology could be used for tasks such as speech recognition or predicting various equations.
TLDR: Researchers from Indiana University Bloomington have developed Bainoware, an artificial intelligence processing unit using brain organoids. These brain cells were trained to predict a non-linear system and showed promising results in terms of learning ability compared to traditional artificial neural network models. This technology has the potential to decrease the time and energy needed for AI training tasks. (Source: Nature Electronics, BioArXiV)
Leave a Comment