The Threat of Artificial Intelligence: Sam Harris vs. Steven Pinker

Last night I watched the Sam Harris “Waking Up Book Club” event with guest Steven Pinker, and I have some thoughts.

First, I admire Pinker as a thinker, and I agree with almost all of his ideas and research about the world getting better despite most people thinking it’s getting worse. The Better Angels of Our Nature and Enlightenment Now are both brilliant books. However, I think he underestimates the potential dangers of artificial intelligence. Like Harris, I am much more concerned about AI.

In the talk, Pinker gave an example that we need not worry about AI deciding to do something foolish like opening the flood gates of a dam because no dam builders would program an AI capable of doing such a foolish thing. He’s right that the dam builders and their programmers would never create an AI that dumb and/or dangerous, but he fails to account for third parties with nefarious intentions.

For instance, what if North Korea creates an AI capable of taking control of a dam in South Korea and opens its flood gates. Or if Russia creates an AI capable of hacking into another country’s nuclear arsenal. Or a million other possible uses of AI with disastrous repercussions.

That’s the real concern. Not general AI going conscious and operating on its own. The real concern is humans consciously using AI for evil.

Pinker also scoffed at the idea of Google programming self-driving cars that would mow down pedestrians. Of course, Google is not who we should be most worried about. It’s governments, militaries, and terrorist organizations weaponizing AI. Or even individual criminals. Someone could hire an AI expert to hack into the self-driving car of his enemy, causing it to crash. Worse still, think of the things ISIS might do with a superintelligent AI.

Evil AI is not the concern. Evil humans with AI is the concern.

When you combine the exponentially improving power of AI with humanity’s propensity for evil, it’s naive not to take the existential threat of AI seriously.

Originally published at https://steemit.com on March 16, 2018.