By imitating the neural networks that are found in the human brain, researchers at the University of Surrey have developed a novel strategy for improving the performance of artificial intelligence (AI). According to a study published in Neurocomputing, mimicking the brain's neural wiring can significantly improve the performance of artificial neural networks used in generative AI and other modern AI models such as ChatGPT.
Topographical sparse mapping connects each neuron only to nearby or related neurons, similar to how the human brain efficiently organizes information. Dr. Dr., a senior lecturer, said, "Our work shows that intelligent systems can be built much more efficiently, cutting energy demands without sacrificing performance." Roman Bauer says that the researchers said that the model had significantly fewer connections that weren't needed, which improved performance over time without affecting accuracy. "Training many of the popular large AI models of today can use more than a million kilowatt-hours of electricity," the doctor asserts. Bauer added. "That simply isn't sustainable at the rate AI continues to grow."
A training method based on biology known as "pruning" is now included in an improved version of Enhanced Topographical Sparse Mapping. This is comparable to the learning process, in which the brain gradually strengthens its neural connections. The research team is also looking into how the method could be used in other contexts, such as more realistic neuromorphic computers, which are computers based on how the brain works and works.
.jpeg)
