> We simply need to make sure that the AGI doesn't optimize anything, but instead takes the saying "all things in moderation" to its heart. That means, every on

> We simply need to make sure that the AGI doesn't optimize anything, but instead takes the saying "all things in moderation" to its heart. That means, every once in a while, AGI should stop whatever goals it pursues and reflect on purpose of those goals, if it is not, by some measure, going too far. A superintelligent AI would be too useful for warfare and making money that we could ever hope to impose limits on what some people will do with it and allow it to optimize.