- Physicists at Leipzig University have created a neural network that uses active colloidal particles for artificial intelligence. - The neural network belongs to the field of physical reservoir computing, which uses the dynamics of physical processes for calculations. - The researchers developed tiny units made of plastic and gold nanoparticles that can process information and make up the reservoir. - The system needs to be trained to perform specific calculations. - The researchers found that using past states of the reservoir can improve computer performance and reduce the impact of noise. - This method can optimize reservoir computation by reducing noise. - The study was published in Nature Communications.