Professor of Astronomy, Harvard University; Director, Harvard Origins of Life Initiative; Author, The Life of Super-Earths
AI is I


Let's take Daniel Gilbert's "end of history illusion," where I think the person I am right now is the person I'll be forever, and apply it to how we think of the human race and our distant future descendants. Our wishful hope for continuity and preserving our identity runs contrary to the realities of our planetary existence. No living species seem to be optimal for survival beyond the natural planetary and stellar timescales. In the astrophysical context of very long time scales, very large space scales, and the current density of energy sources, our biological brains and bodies have limitations that we are already approaching on this planet.

If our future is to be long and prosperous, then we need to develop artificial intelligence systems, in the hope to transcend the planetary lifecycles in some sort of hybrid form of biology and machine. So, to me, in the long term, there is no question of us versus them.

And in the short term, the engineering effort to develop a more capable AI is already producing systems that are left in control of real-life stuff. The systems fail sometimes, and we learn of some of AI's pitfalls. It is a slow and deliberate process of learning and incremental improvements. This is in contrast to discoveries in science, when new physics, or new biochemistry could bring about a significant engineering breakthrough literally overnight. If the development of AI is less like a phase transition, and more like evolution, then it would be easy for us to avoid pitfalls.

After almost 4 billion years the ancient poster children of Earth life—the microbes, still rule the planet. But the microbes have no exit plan when the sun dies. We do, and we might just give them a ride. After all, those microbes may still be closer to our present selves—representatives of life's First Generation rooted in the geochemistry of planet Earth.