Skip to content

Latest commit

 

History

History
27 lines (12 loc) · 1.65 KB

No.Neuron.Learning.Machine.md

File metadata and controls

27 lines (12 loc) · 1.65 KB

Artificial-intelligence system design ideas

A learning program design that doesn not use a neuron

Introduction

An AI uses neuron concept that doesn't make any sense. Human neuron is constant learning, editing DNA cycle where as AI neuron is out of nowhere put in a computer.

Current system

Artificial learning models are based on neurons. The neurons as we use are fundamentally flawed as they rely on threshold, which seems too artificial and does not reflect human neurons, hence it is not a good learning agent. It is also flawed because neuron in human shares biology with its surroundings, whereas artificial neurons does not share any commonality in the environment it runs. In other words an AI neuron feels like it does not belong in the place it should.

Correct and proposed system

Computers run program. Program lives and dies by the parent process that spawns it, in a way it reflects neuron's life cycle. The learning agent is a program that is called, executes, checkes its correctness and edits itself. The cycle continues untill it learns about itself and its environment.

System design

Conclusion

Artificial Neurons are well marketed, not well thought through concept somehow came into trend. It sounds nice but it doesn't fit very well. Neuron is a biological part of a living organism. If a learning agent is required for a computing device, it should be a part of itself, which is an executable program. Neurons do not learn, Its DNA that learns. Program is a DNA of a computing machine, NOT NEURON. Live, Learn, Die. Run, Edit, Kill.