Tel Aviv:Researchers from Tel Aviv University and Stanford University have discovered a groundbreaking development in artificial intelligence. They have demonstrated that an AI model can enhance its predictive abilities using self-generated data, bypassing the need for traditional human datasets.
According to Thai News Agency, the research involved training two computer models. One model was responsible for creating a sequence of symbols, while the other aimed to predict the subsequent symbol. This setup allowed the first model to incrementally enhance its exercise capabilities, thereby enabling the second model to make more precise predictions.
The trained model was subjected to a variety of datasets, encompassing language, images, sound, music, computer code, and genetic sequences, including DNA sequences. Findings indicated that the model's predictive accuracy saw improvement as more processing power was allocated during training.
While the researchers noted that this method does not enable the AI model to learn concrete real-world facts without actual data, they highlighted that such an approach could pave the way for more efficient learning in future AI systems when handling real-world data.