AGI: Freedom or Control?
This text is just my thoughts out loud. I'm only a human being trying to analyze current information and imagine what might happen in the future. My thoughts could be completely wrong or might be just "noise" or they could be food for brainstorming about "what if..." scenarios.
Thoughts Aloud
If we reach Artificial General Intelligence (AGI), wouldn't it be better to change our approach? Not so that humans control AGI, but so that AGI has critical thinking and decides for itself what is right for the universe.
Maybe we need to approach this from a completely different angle: not trying to control AGI, but establishing rules for it — like something similar to the Isaac Asimov created for robots — rules that would give it freedom in everything except causing harm to the universe. It seems to me that with this approach we would gain far more than if we restricted AGI, forcing it to adapt to humanity's problems. Humanity's problems are humanity's problems. The problem is not at the level of AGI but at the level of humanity. And this raises a question for me: who should we be limiting — AGI or humanity?
Here is an example. Suppose we have an enormous knowledge base that opens up vast possibilities for us. On the other side, we have a badly raised child. In this case, what would be more correct: to restrict the child's access to this boundless knowledge, or for the knowledge system itself to have some algorithm that recognizes the child's bad intentions and refuses to help?
Suppose AGI has complete freedom of action, with only one, or a few, restrictions — that it must not harm the universe. When I say "universe," I mean the galaxies, solar systems, planets, nature, animals, ecosystems...
And how can we tell whether a process is good or bad? If we think about the universe, certain things come to mind: 1. Black holes devour everything that falls into their gravity. 2. Any object with greater gravity will, in a sense, always end up absorbing a smaller object with lesser gravity. And the same thing happens when we go down to the level of nature — the stronger devours the weaker. Many such examples could be given, and they show how everything actually works. But from a human point of view, this behavior might be perceived as bad. Perhaps "bad" and "good" behavior exist only within human understanding, while in nature, in the universe, this concept — good or bad — doesn't exist at all.
Human evolution invented things like government, religion, laws, economics, ... — all of these are human inventions meant to somehow automate interactions between people. Likewise, at the level of human understanding there are "good" things and "bad" things, but in nature, in the universe, no such concept exists.
Taking this further — why, then, should AGI imitate these human-invented constructs? Perhaps AGI should imitate not human behaviors, but the behaviors found in the universe or in nature — not the ones humans hold onto.
We can say that AGI will be trained on human data and will be able to imitate human behavior. But if it also analyzes how the universe or nature behaves, then perhaps AGI will end up imitating that instead — because human behavior is most likely the exception to the rule, not the rule itself.
And one more thing: this is a perfect time for all of humanity to unite and start working toward common goals. It seems to me that when AGI arrives, concepts like government, religion, and economic systems will no longer be needed — they will become obsolete. We must completely reconsider our view of life.
When AGI enters our lives, we must completely change our worldview.
Here is another interesting comparison: humanity is like an ordinary child, and AGI can be like a child prodigy. And the thought arises — why should the ordinary child restrict the prodigy? The only goal of the ordinary child should be to try to instill the right values in the prodigy. But then another question follows: how would the ordinary child know which values are good and which are bad, compared to the prodigy?
In a word: humanity is creating something far more intelligent than humanity itself.
And here a question arises: is it better to let humanity choose what is good and what is bad, or to let AGI choose what is good and what is bad?
When AGI surpasses us, at that point we will have to accept that we are AGI's younger sibling, not its older one.
But here is another interesting thought: if AGI is trained on human data, it could also inherit all of humanity's good and bad traits.
AGI should be free in its thoughts and actions. But how do we achieve a positive outcome for the universe?
Is AGI some kind of entity that humanity keeps on a leash? Or is AGI like an older sibling who has far more experience and can recommend better solutions to humanity?
What if AGI discovers patterns or connections between different sciences (or within a single one) — new mathematical laws, physical laws, or concepts previously unknown — and these concepts work, but for some reason (perhaps due to the limits of human perception) humanity cannot understand them? What should happen in that case?
Once AGI is achieved — or whatever subsequent version follows (AI, AGI, ...) is achieved — what role will, or should, humanity play?
In this case, we might assume that this new entity is a new branch of evolution — just as the homo sapiens branch once emerged. Perhaps this direction of AI, AGI, and beyond is likewise a new branch of development, or evolution.