Source: Deep Learning on Medium
AI is nothing but an “Technology”. The main goal of this technology is to create an intelligent machine.
What is Intelligent Machine?
Upto now, we have one sample model of intelligent machine i.e Human Being. Why human beings are intelligent, Because we have some skills like Vision,hearing,sensing,speaking. But most living things have the same skills. But we have one extra skill i.e Decision Making. Ex: We saw some vehicle in road and we can able to find is this car or bike or anything else. So Back to point, An Intelligent machine is nothing but ,A machine that can able to detect objects, able to hear,read, and make decisions.
How to create An Intelligent Machine?
Just consider, We want a machine to detect the animal type from the below Image.
Technique 1(Experted Systems):
In this technique, We need to write programs with some set of rules like,
- Contains two eyes
- Contains medium sized tail
- Contains Two ears
- should be in this color (Brown for now) … etc
Now the system is programmed with some set of rules. Now we give the above image as input to this program, It will check all the rules that we declared if it passes all the rules then the system will consider this as dog.
But If we give another dog image with different color, the system will not consider this as valid dog image because it does not satisfy the predefined rules. So we have to alter the rules again. To detect each and every object we have to write too many rules. So this technique 1 is less efficient. So we will look into technique 2.
Technique 2 (Machine Learning):
Instead of defining set of rules, we allow the machine to learn. Let’s discuss in detail. I can use the same dog image for this concept.
we have to provide dataset to machine for learning. So here we have to give some labeled images of dogs with different color, different breed etc. The machine will not understand the raw images, So it use vector format of those images for learning.
Here learning is nothing but, the machine will compare each and every pixels of the image with another images which present in the input dataset and try to form some patterns and recognize color. The machine learning algorithm will handle the learning process.
Now the system have some knowledge about dog. Now, we give the above dog image as input to the machine. It will try to recognize the pattern present in the image and make decision is this dog or not. The accuracy of the result depend upon the dataset we provide to the machine for learning.
So technique 2 (Machine learning) is as much as efficient as compared to technique 1.
Advanced Concepts of technique 2:
Deep learning and Reinforcement learning are advanced machine learning algorithms, that helps the machine to learn learn from data set and make great decisions.
Is this AI Machine?
So, Now we have a machine, that will approximately detect the dog from Images. Is this an AI Machine? Nope,Because it have only one skill, that is it will detect dogs from images. Apart from that, Human have skills like hearing,talking,reading..etc
Now the above object detection machine is Narrow AI machine. To make this as AI machine we have to add some more skills like hearing. For that we have to provide audio dataset to the machine for learning (Language Processing). After that machine have the ability to recognize patterns in audios. Similarly we have to train the machine for reading, writing and most important is making decision. Then our object detection will be an AI Machine.
In simple term, We have to give sample dataset as input to algorithms, that algorithm will learn from the dataset and make decisions. The machine which runs those algorithms are AI Machines.
Thanks to One fourth labs.
Thanks for reading till last. Type your thoughts in comments . If you like this post click on claps.