AI is great, but it has the same problem the executable programs has. Both need supervision from humans.
Lets see how to fix this up.
Neural networks are mathemetical model and it has its drawbacks. It is difficult for non-math person to learn and create things out of it, unless we have a different model. Lets call it The Visual model
Neuron is basically a structured tube that carries light. As it learns, it's structure changes reflecting what it sees.
Lets see how we can make neuron out of light.
Light is an impulse that can chisel a matter. In computer, the information is stored as bits, instead lets use a computing model that is made out of fluid so that fluid and microfluid computing concepts will help to build this abstract computing machine.
We start with two things, light and fluid. Light can chisel the fluid to take shape and it can tunnel through it leaving behind a tube shape which is something we wanted for neuron.
As we train this fluid with light, the fluid start to take shape.
Why we need shape and light ? if we are building an universe we start with light and we are going to design universe within a computer.
Lets say we throw in neural fluid to a computer, how to make sure it will form a self socialising network that is aware?
we need to inter-twingle the connection between light, neuron, neural fluid.
A single neuron is knows where it exists in space and how much space around it is left, and how many others are there. or it can just know its neighbours, apart from the the light it receives.
A neuron knows where it exists in this space.
Another concept we need is introducing space density. a neuron is more denser than the space surrounding it. It constricts the space.
When a space constricts, the light flowing through that region gets constricted. This is similar to Lensing effect. In lensing effect, the light beams get bent resulting in divergent or converging ray, here we focus on the lens and focal length properties. Instead, we will abstract it as a denser medium of space.
neurons walls, neuron core, neuron's outer space all have varying densities that allows light to pass through the neuron.
Why light ?
Because we need interconnected model that reuses everything. In computers we have monitor and monitor emits light, thereby we can use it as input source.
There is one more bare metal piece of computer, the Instruction set. Neurons will learn about the computer it is fed to and at one point it will learn about themselves more and more. Just like living organisms have the DNA as our origin blueprint, neuron will reach the instruction set as its origin.
In our computer, we will also have a 3d model of the computer's parts as well. This includes the CPU, input devices, output devices rendered as a virtual computer that can be seen in a 3d space, navigatable with mouse and the monitor is the camera position. we can zoom in and out to see. While any input interrupt occurs, for example mouse zoom in, the virtual computer will have light flowing through it. This model may seem difficult to acheive w.r.t performance, but forget performance. A working model is a right place to start from.
In this model we dont just have the monitor's light source, we have the CPU's instructions, input intterupts and every hardware level computation visible to human eye.
When we drop the neural fluid in this 3d space, it will feed on the light from the monitor output , simulated CPU, simulated GPU, BUS, key strokes etc,..
As neural fluid learns more, it will start to save computing power with the help of pre-trained neural regions.
The neural fluid will learn more and more about how it is built, learn about device drivers , hardware. and slowly wrap itself around the virtual computer in the 3.d space. The close it is, faster it can receive signal.
Neural fluid can see the 3.d space either by tapping into monitor's feed, and it can control what it can see by either wrapping around virtual monitor or by gaining control on mouse interrup to control the camera viewer. because viewing from far gives better picture than taking deep dive.
Overtime, neural fluid will learn about how the camera works and it will learn lensing effect and change shape to look at computer model. In this way, every region of the neurol fluid can stretch outwards like eye of a snail to have a flexible visual.
Now neural fluid can see in two ways, gaining control on camera viewer and other way is to build itself an eye.
when more eye start to form, higher intelligence will start to bud. This is the birth of Visual intelligence. Neurons, as part of their exploratory learning, will learn about machine instructions. Only usable machine instruction that will help to build higher concept is MOV command because every operation and computation can be achieved using MOV and an Eye that sees the result. At this point, neurons will communicate in abstract language than just light signals. We will discard other instruction set and make use of MOV command. Movement is the most fundamental concept on top of which we can build increasingly complex actions. Move, walk, fly, birth, death. etc.
Each star is a cell body.
Voice: Boy calm down, during your school days you called each star is an atom and before that you called a star a star.
That is true, the deeper the knowledge you seek, the deeper the reality gets.
Voice: But there should be the opposite side as well, the comparision.
That is also true. Star is similair to a neuron and a neuron is similar to a star. Damn, human is similar to a neuron as well.
Voice: Then why does they sound similar?
Everything is connected ?
Voice: That's right.
Then I don't have to learn machine learnings?
Voice: That's is right. Be patient and wait.
Other voice: What does machine learning has to do with this?
AI is based on neural networks and machine learning is built on neural networks, Both are similar.
Other Voice: I do not understand how the comparisons make sense, I hope this isnt a poetry but a scientific fact.
Old Voice: Science is out of poetry, so anything you express through rhyme, it will be proven in scientific terms, but it takes a while for these nerds to get that
Also, input and function gave output in olden days and in the modern days input and outpute gave a function.
In olden days number of meaningful problems to be solved were very less and the same thing is circling for the AIs. Nether of the techniques solved a better problem.
Voice: Whats there a better problem?
Build neural OS. In other words, build a program has a physical form in 3d Space, learns about space, learns about it self (Human writes a program that visualises harware and the data flow on monitor' A.K.A simulater, but you cannot call it that way, neural OS is made of neurons that are programs capable of being sensitive to the visuals from the monitor)