Today I wanted to share feelings and thoughts I often have while listening to discussions about Artificial Intelligence.
We have heard a lot about this field, what it is doing now and its promises. We have also heard a lot about expectations and fears for the future. And it is not going away. 🤖
My main feeling is that we are still in that point in our discussions where our imagination is too limited to envision true potentials and weirdnesses 👽 of AIs.
The discussion is always monolithic. Filled with anthropomorphisation and historical comparisons. Here are some things I think we ought to spend more time talking about in AI.
Consciousness and ethnocentrism
Being able to produce new forms of intelligence that:
- count faster than we do.
- process data faster than we do.
- understand images as well
- speak like us.
- create and imagine speech, images, …
Has required us to constantly change our definition of intelligence but also of consciousness.
While it has no clear scientific meaning or definition 📖 and while this list continues to expand. We might have to let go of this adjective/noun. Consciousness is certainly just a figment of our imagination 🤷. A feeling that we associate to things that our minds try to emulate, things that we judge similar to us. Through History, Humans have deemed conscious sometimes only small specific groups of other humans, like white europeans, believers… sometimes animals, rivers, forests and trees (through spirits and divinities).
Even worse, this acceptance has to make us rethink our notion of self altogether. This too has been relegated to a likely… mere illusion.
AI debates could go into so much depth if it dared do what it should. Help us put into question our “still” very strong ethnocentrism and reevaluate some of our most strongly held beliefs. Who knows? It might even allow us to create a better society.
Intelligence at least does not seem to be an illusion and it is a fairly rich scientific subject. We might be able to better understand and define this illusion we feel as “consciousness” and intelligence research might be the right way of doing this.
It too seems to lie on a continuum. It seems to be related to the ability of an agent to have some specific form of information processing and some interaction with the world, both by receiving rich inputs and by being able to generate complex outputs to this world. 🏛️
We now know that raw computing power is not intelligence. It is only its substrate. A super-computer only matters in regard to the algorithms it is able to perform. Interaction with an environment seems also greatly important in creating intelligent machines. 💻
By this logic we understand that human beings are not born intelligent. They are only born with the substrate and the algorithm to create intelligence out of interaction with its environment.
what intelligence can be
can we see extremely different intelligences or do they revolve around the same points?
A clear difference exists between a computer and a human being’s brain. A brain 🧠 has its algorithm encoded in its substrate. whereas a computer’s CPU is open to any possible computation and is completely separate from the computation it performs.
So a computer can compute anything. We can visualize this “everything” with a circle. 🟡 A human brain can do a very small subset of the computation. It is specialized. an infinitely smaller circle within this big outer circle. All possible brain configurations we have ever seen only represent a bigger but infinitely small circle in the space of possible computation. 🟠
The question “what intelligence can be?” could also be laid out as: “How big the possible-intelligences circle can be?”
From what we can sample: intelligences on earth. It seems that intelligence can be different enough from ours that we have difficulty recognizing it. It seems that animals that can reproduce many acts of intelligence: math, reasoning, creating counterfactuals, etc. Although most people would not put them in the same realms as humans as we have complex social behaviors, culture and complex speech. 😒
I think it shows that intelligence can be different enough to not be recognized by the naked eye. And I believe that to answer that question of “what intelligence can be?” and because it is only fabricated out of its interaction with its environment we will need to first know “in what breadth of environments can it flourish?”
what will we see?
I don’t know. 😛 Crazy things I am sure. 😏 I will only answer this with an advice. Do not put too much weight on your view of the world and how it works. And always be on the lookout for a better conception of life. These inner axioms of ours are most surely imperfect. They will lead to nonsensical reasoning. Like christianist’s axioms that will lead to statements like “but if god is good why is there bad?”, “if my neighbors sinned why did he received a promotion?”.. The answer to all those nonsensical reasoning is that their axioms are fraudulent. When such things arise from a set of axioms, it means that the set of axioms is not good enough anymore to reason about the world.
Let our discussions about AI stop be “oh damn it is going to take control”; “will it enslave us”; etc. Let them be about ourselves because as we will get increasingly better AIs they will shade lights 🔦 on questions we could not have imagined and destroy the strongest assumptions we held about ourselves. Will we be prepared? As a society, can we find the axioms we need to understand, accept and control it?