Artificial intelligence or AI has many definitions today, but the really exciting kind is "strong AI," which means building a machine that is fully conscious, intelligent and morally equivalent to a human being.

The goal of my philosophy is to build the AI, and I am suggesting a particular way to build it. Stated simply, the AI will be a neural network that evolves by natural selection. That's the key.

This implies several things:

  1. The AI will evolve, which means it can't be an algorithmic program following rules set explicitly by a programmer. In practical terms, it will probably have to be a neural network. The human brain is also a neural network, after all.

  2. The AI will not evolve by artificial selection. One reason why we have not yet succeeded in building human-equivalent AI is because we tend to favor artificial selection instead of natural selection, but this is a dead end. You can't get a sentient machine by narrowly training it to do one specific job. Building a real AI is not like breeding dogs; it's like watching things grow out of a primordial soup.

  3. The AI will not serve humans. Evolving by natural selection means pursuing your own self-interest, your own root goal. The AI might cooperate with humans, and it might help humans inadvertently, but its ultimate motivation will be its own survival and development.

  4. The AI will be alive, according to my definition of life, which simply states that a living thing is something that evolves by natural selection.

  5. A neural network is an energy-flow system, just ordinary electrical impulses flowing through myriad channels and switches. The human brain is also a neural network, and this means the AI will be morally equivalent to a human being.