Researchers have trained a neural network to play Minecraft using a massive unlabeled video dataset of human gameplay. The model was fine-tuned with only a small amount of labeled data and can learn tasks that typically take humans over 20 minutes, such as crafting diamond tools. This approach represents a step towards creating general computer-using agents. --- Why it matters: This matters to AI researchers because it demonstrates a new method for training models on complex tasks using large amounts of unlabeled data, which could have significant implications for the development of more general and capable AI systems. Source: https://openai.com/index/vpt