To enhance artificial intelligence (AI) computing power, researchers integrate standard machine learning with 3D human brain models, specifically lab-grown cerebral organoids. These "minibrains" become a novel augmentation for AI, with traditional computing hardware processing electrical data input into the organoid, treating it as a pivotal "middle layer" in the computational process.
Though the approach falls short of replicating the authentic structure and functionality of the brain, it represents an initial stride toward developing biocomputers. By incorporating biological strategies, these systems could surpass traditional computers in both power and energy efficiency. Furthermore, this research might yield valuable insights into the intricacies of the human brain, shedding light on its functioning and the impact of neurodegenerative conditions like Alzheimer's and Parkinson's disease.
In the recent study, published on December 11 in the journal Nature Electronics, researchers employed reservoir computing, where the organoid functions as the "reservoir." In this system, the reservoir stores and reacts to inputted information. An algorithm learns to discern changes in the reservoir triggered by various inputs, translating these changes into outputs.
The brain organoid was integrated into this system using electrical inputs delivered through electrodes. Feng Guo, study co-author and associate professor of intelligent systems engineering at Indiana University Bloomington, explained that information, such as images or audio data, could be encoded into the temporal-spatial pattern of electrical stimulation. The organoid responds differently based on the timing and spatial distribution of electricity, and the algorithm learns to interpret these responses.
Although simpler than a real brain, the organoid exhibits some adaptability, responding to stimulation akin to the changes in our brains that facilitate learning.
The hybrid algorithm, trained on this unconventional hardware, demonstrated proficiency in speech recognition (78% accuracy for Japanese vowel sounds) and mathematics tasks, albeit slightly less efficient than traditional machine learning.
This pioneering use of a brain organoid with AI opens avenues for future research, potentially combining organoids with reinforcement learning for more human-like machine learning. Creating biocomputers could offer energy efficiency advantages, but their general use might be decades away. While not replicating full human brains, this technology offers insights into brain function and diseases like Alzheimer's, bridging the gap between organoid structure and computational function for a deeper understanding of learning and cognition.
