Technology, Artificial IntelligenceJuly 17, 2023Human-Aware AI accelerates scientific discoveries: A paradigm shift in research capabilitiesA recent study published in Nature Human Behaviour explores the potential of human-aware artificial intelligence (AI) to revolutionize scientific discoveries. The researchers developed advanced models capable of predicting human inferences and generating novel hypotheses that may not be considered for decades. By incorporating AI that is aware of human expertise, the study demonstrates the ability to accelerate scientific progress by surpassing human limitations and complementing human capacity. The models, trained on published scientific findings, outperformed traditional AI models by considering the distribution of human scientists involved in research. Through extensive simulations and random walks across research literature, the models exhibited a remarkable 400% improvement in predicting future discoveries, especially when relevant literature was scarce. Furthermore, the models achieved over 40% precision in identifying the scientists who would make those discoveries, illustrating their understanding of the connections between individuals, experiences, and the properties being investigated. The lead author describes the model as a "digital double" of the scientific system, enabling simulations and experimentation of alternative possibilities. The study also conducted a demonstration where the AI model generated scientifically plausible predictions least likely to be discovered by humans, exploring entirely new territory. By reframing AI as radically augmented intelligence rather than artificial intelligence, the study emphasizes the importance of understanding human cognition and designing systems that compensate for its limitations. Ultimately, AI has the potential to push scientific progress beyond the boundaries of current knowledge by expanding human capacity and supporting improved exploration.
Source: Technical University of Munich (TUM)