In a new research published in the journal Nature · Machine Intelligence, scientists from the Netherlands National Institute of Mathematics and Computer Science (CWI) showed how brain like neurons can be combined with novel learning methods to train fast and energy-saving spike neural networks on a large scale. Potential applications include wearable artificial intelligence (AI), speech recognition, augmented reality, and many other fields. This spike neural network can be implemented on a chip called neuromorphic hardware, which is expected to make AI programs more user-friendly. This solution is beneficial for protecting privacy, improving robustness and responsiveness, and its application range ranges from voice recognition in appliances, healthcare monitoring, drone navigation, to local monitoring devices. Just like standard artificial neural networks, spike neural networks also require training to smoothly perform these tasks. However, this type of network communication also poses serious training challenges, as they cannot be compared to the learning ability of the human brain: the brain can easily learn from new experiences, change connections, and even establish new connections; The brain requires very few 'templates', but learns a lot; The brain is also very energy-efficient when learning new things. In order to achieve a level of similarity to the human brain, new online learning algorithms can directly learn from data, achieving larger peak neural networks. In the researchers' presentation, the underlying spike neural network SPYv4 was trained to distinguish cyclists, pedestrians, and cars on a busy street in Amsterdam, and accurately indicate their positions. Researchers have stated that in the past, they could train neural networks with over 10000 neurons; Now, for networks with over 6 million neurons, they can also be easily trained. With powerful AI solutions based on spike neural networks, researchers are developing chips that can run these artificial intelligence programs at very low power, and these chips will eventually appear in many intelligent devices, such as hearing aids and augmented/virtual reality glasses. Modern artificial neural networks are the backbone of the current AI revolution, but they are actually products inspired by real biological neural networks such as the human brain. Indeed, the brain is currently unmatched by any AI - it has a larger network, works more energy-efficient, and can respond faster when triggered by external events. How to get closer to the real brain? That's the work of mimicking biological neurons more realistically. Scientists have found that neurons in the human nervous system communicate by exchanging electrical pulses, and spike neural networks have become a special type of neural network in this study by imitating this. (New News Agency)