The world is changing at an increasingly rapid pace. In the span of only a few short years, we have witnessed technological progress, population growth, and globalization to a degree not seen in the lifetimes of our ancestors. Machines are also becoming smarter and more capable. As automation increases in domains where human labor and decision-making were once necessary, it will become increasingly difficult for individuals to create value and meaning through work. And, if one does find a niche, further changes may soon take place—new technology will acquire new skills, and people will continuously need to adapt. As a consequence of this growing dynamism, it is no longer sufficient to adapt to any one environment; humans and society must learn to adapt to change itself—they must increasingly learn to learn. In this paper, we begin with a brief account of how brains and minds work based on a theory broadly known as predictive processing (Friston, 2003; Clark, 2013). According to this view, humans come to understand and perceive the world by making predictions, a process that is therefore at the heart of understanding how humans deal with unpredictable circumstances. We then discuss research on how humans and machines respond in situations characterised by volatility, uncertainty, complexity, and ambiguity (VUCA), and the role of agency in social and moral situations. We conclude by arguing that learning-to-learn and meta- learning strategies are key capacities that currently distinguish humans from machines. For society to be generally adaptable to change, we propose that social structures and education systems will need to nurture skills that foster general and transferable learning capacities (rather than, or in addition to, specific skills). For humans to flourish in the future, governments are also encouraged to incentivize citizens who possess skills to become teachers and mentors. Society can be made robust when experts are inclined to teach those who are willing and able to learn.