Summary: Scientists have developed a neural network that exhibits human-like systematic generalization in language tasks. This AI system can incorporate newly learned words into an existing vocabulary and use them in various contexts, mirroring human cognitive abilities. The research, published in Nature, suggests a breakthrough in training networks for systematic reasoning. This development may lead to AI systems that interact with people more naturally, addressing the gaps and inconsistencies present in current language models. The study could potentially enable more efficient learning in neural networks, reducing the need for vast training data and minimizing errors like hallucinations.