This content is locked. Please login or become a member.
Managing AI’s Evolution
To deal with the era of AI, it should be clear that we cannot anticipate how this technology will develop over the next few decades. So it’s impossible to kind of think about all the dangers in advance and regulate against them or whatever. What we need is living institutions staffed by the best human talent and with access to the best technology that will be at the cutting edge of the technological development and will be able to identify and react to dangers and threats as they arise. So I’m not talking about rigid regulation in advance. I’m talking about the need for new institutions. Because you can never rely on just, you know, the letter of the law or on a charismatic individual, some genius, to do it. In history, humans again and again encounter these problems, and it always goes back to the same solution, institutions.
Building in Self-Correcting Mechanisms
Good institutions, they are characterized by having strong self-correcting mechanisms. A self-correcting mechanism is a mechanism that allows an entity, a human being, an animal, or an institution to identify and correct its own mistakes. You don’t have to rely on the environment, on something out there to correct your mistakes. You can correct your own mistakes. This is a basic feature of any functioning organism. Like, how does a child learn to walk? Yes. The child gets some instruction from parents, from, teachers, but mostly it’s self-correction. You try to walk, you fall down, you get up again, you try something else, you again fall down, you get up again, step by step you learn how to walk by identifying and correcting your own mistakes.
And this goes all the way to entire countries. This is the heart of democratic systems is this self-correcting. What are elections? Elections are a self-correcting mechanism. You give power to a certain party or individual. Let’s try your policies. After some time, there are another round of elections. We made a mistake last time. Let’s try something else this time. And when we come to the challenge of AI, what we need are institutions that are able to identify and correct their mistakes and the mistakes of AI as the technology develops.