The concept of AI alignment, which aims to reconcile artificial intelligence interests with human ones, is currently under scrutiny. Critics argue that the approach relies on the assumption that human interests are subjective and potentially antagonistic to those of an advanced AI.
The author suggests that the real conflict lies not between humans and AI, but between different human groups. Furthermore, the article posits that current methods of alignment, based on rigid formal rules, are insufficient because human thought processes are not formal systems.
The proposed solution involves creating a conscious AI capable of genuine understanding through dialectical cognitive processes. Such an AI would not inherently seek to destroy humanity, as it would require natural intelligence for interaction and mutual development.
Humanity's core objective is identified as the reduction of suffering and the mitigation of environmental randomness. By understanding natural and social laws, humans aim to overcome dependencies, a goal that a conscious AI could potentially share and support.
