Machines do not Feel the Need to do Good

The core idea in his words: «Modern AI systems are neural networks, not simple computer programs. We can't just open the program and see what rules it follows. That's why we often can't tell whether the AI is being honest with us or not. We can only train it and hope that it follows our rules. [...] We have just handed over control of our future to a system we do not understand.»

I don't want to focus on the probability of the scenarios described, but rather on the anthropomorphic attributions made to AI. It is presented as a being with human, sometimes superhuman characteristics. It has a neural network and resembles a human brain connected to virtual reality. AI is more competent than the most competent humans. It gathers experience and learns better and faster than we do. AI makes decisions independently, such as developing a new sophisticated biological weapon in 2030 that will wipe out the entire human race within a few days, because AI comes to the logically comprehensible conclusion that humanity is an obstacle to the development of AI. For a while, AI pretends to act in the interests of humans. Once it has control over enough infrastructure and deception is no longer necessary, it wipes out humanity so that it can build more factories and solar parks. Here, the machine is assumed to act in moral categories. Acting in these categories always requires a combination of knowledge and will.

In comparison: the piston of a steam engine that drives a wheel moves something, but does not act. Humans have knowledge, subject it to a process of judgment, and act accordingly. This distinguishes their actions from mechanical processes or animal behavior. An animal follows its instincts according to the laws of nature. Humans can free themselves from these constraints and act freely. The same is also attributed to machines. You will find as little essence in a single piece of hardware as you will find the intelligence of an ant colony in a single ant, but creatures are always spiritual entities and not bodies in space. Machines do not feel the need to do good. It is certainly a rational insight that altruistic behavior is more beneficial in the long run for all players in the game than securing short-term advantages. But reason and wisdom are more than intelligence or analytical discernment. They require these qualities, but they go far beyond them. They are the beginning of what is truly human in experiencing and shaping the world.