Artificial Intelligence is a field that wants to make intelligently acting machines.
Where do we see intelligence in nature? Technically everywhere because it seems to be a manifestation of atomic particles but well expressed in living organisms, most in humans.
How are people working on this goals right now? There are “frameworks” which are libraries of programs. Intelligent systems are then written using them which has a basic core structure like so: The libraries have certain functions that allows to act on input for example image manipulations. System has a few mathematical functions whose parameters are adjusted to fit the desired outcomes by providing the system with inputs which are labeled for outcomes. After adjusting the parameters, the system is put to check if it works with unlabeled inputs. If it does, it is a good intelligent system. If it doesn’t, loop.
One might argue, that these libraries are designed by keeping in mind the theories of intelligence. But that is not technically so. Having multiple layers of a mathematical functions doesn’t mean it is a neural network unless it stimulates what neurons do - because that is how the name Neural Networks was coined.
So, what are they doing? They are playing with mathematical functions, and applying them on crudely defined architectures.
One might argue that they’re mathematical functions, at least they are well understood. Well, no. Equations of calculus were derived not by randomly trying out things; they were derived by applying intelligent thoughts. Calculus was not a trial and error derivation, Leibniz (or Newton) worked hard to understand how things work, and proposed a mathematical model that explained the observed behaviors well. So, a correct approach towards AI would have been to look at the behavior of nature, understand it well and propose a model that tries to explain the behavior (or a part of it).
This seemingly ridiculous approach to the field doesn’t even look like Science.
And this makes me think that a very high percentage of the “community” is doing something I do not understand. That is not to say that those randomly (pseudo-randomly) trying out things are fooled. It’s like computer hackers trying out things to see what does what. These hackers have been known to find out amazing things!
The very popular book on AI by Russel and Norvig talks about Searching algorithms including heuristics. This might give an impression that the experts consider heuristic search approaches as a part of AI. But they make it clear, that these approaches are the bases on which the AI stepped to gain its current step. On the other hand, Marvin Minsky suggested in one of his lectures on The Society of Mind at MIT that he skimmed through the book and “could not see any AI”. (Minsky suggested Artificial Intelligence by Patrick Winston instead).
It is also not to be confused that such an approach is not useful. It in fact is very useful. For example, a linguist friend of mine pointed out that many in the field of linguistics use these tools to help them find out structural aspects of the languages - for example by speech segmentation. This shows that the non-intelligent architectures are indeed helpful in finding out how intelligence works. But I do point out that the tools being used are not intelligent and have very little to do with AI.
A note to be taken is that I am no expert and do not claim to know enough to attempt serious debates over the field of AI. I have heard those experts with whom my ideas seemed to resonate and have agreed to their ideas. I understand that there are many who are not following the claimed false approach which makes me very happy.
I understand that since I am on the wagon of saying “this approach is wrong”, I should speak of which approaches are good in my view. I will be talking about that in another post.