20 years from now, I suspect we’ll look back at some of today’s AI debate and ask ourselves: why were we so afraid of AI?
Not because the risks weren’t real. Every transformative technology creates new risks, and AI certainly does. And yet, we have always learned to adapt to powerful new tools, building norms, institutions, and safeguards along the way. What troubles me more is that fear is becoming the dominant lens through which we understand AI.
Over the last few years, public discourse from some frontier AI labs has over-emphasizes catastrophic risks, national security threats, geopolitical competition, and worst-case long-tail scenarios. And mainstream media has heavily leaned into these narratives. Alongside massive capital investment, this framing has often become the dominant angle. Stepping back, it often feels like the way we talk about one of the most important technological advances of our time is increasingly synonymous with danger.
Meanwhile, we don’t talk enough about the use of AI to discover new drugs, accelerate scientific research, improve healthcare outcomes, make education more widely accessible, design new materials, modernize physical industries, expand access to expertise, etc. These are incredibly hard problems, and thousands of engineers, scientists, and entrepreneurs are working on them from different angles. Many of these gains will likely take time to materialize.
The remarkable thing about AI is that it is fundamentally a net-positive-sum technology. As intelligence becomes more abundant, more accessible, more distributed, and cheaper, entirely new opportunities emerge. The pie itself grows.
My optimistic vision for AI is not one where humans become less important. It’s one where we spend less time on routine cognitive work, and more time on what makes us human: thinking originally, building communities and nurturing relationships, creating for the joy of creation, exploring weird non-economically-viable ideas, etc.
I suspect the long-term impact of AI won’t be determined by the risks we are imagining today. It will be determined by what we build once intelligence becomes abundant, distributed, and widely accessible.
