Part 3/10:
The motivation for exploring unsupervised learning lies in the complexity of real-world tasks that intelligent systems face. Supervised learning often relies on abundant labeled datasets, which can be difficult to obtain. In contrast, unsupervised learning draws on the wealth of unlabelled data available in our digital world, mimicking human learning more closely.
As humans learn primarily through exploration and observation rather than direct instruction, it is believed that systems that adopt similar methodologies will achieve more sophisticated forms of intelligence. Furthermore, enhancing generalization capabilities to adapt to new tasks and situations—a cornerstone of intelligence—can be problematic in purely supervised or reinforcement frameworks.