Why Machine learning for achieving Artificial Intelligence? “ The Need for Machine Learning “
Even though AI can be achieved in many ways why does machine learning has more edge over others? ( A must read for beginners ) Why is it called Machine Learning?
I don’t believe in blindly learning things so we will start off with why everyone is using machine learning in AI for better understanding.
Now let us assume we can code for all the properties in identifying fruit where each function contains properties listed as in traditional programming.

Go on, list the properties out. Some of the properties listed commonly are as follows:
- Color: “Red” — “Orange”
- Shape: “Not completely circle” — “circle”
- “Stalk is present” — “Stalk is not present”
and much more properties right??(if you have more properties list them in your mind or on paper for now)
Now consider if the image below, what if the image is black and white???(the first property listed is useless now)

Let us consider only apple now.

Easily identifiable right??? APPLE 100%
Not soo fast…

So all the properties listed by you are any of them valid now???
So it is impossible or in other words, it is impractical to list out all the properties in all the possible dimensions of each and everything that needs to be identified.
Some people may raise questions like these are too similar to classify give us a much easy classification task…
Okay check this out…
Quite easy to classify right???



Now do we all agree that there is a need for much better methods than traditional properties listings in functions So now the need arises for a different method of coding and we have Machine Learning to our rescue.

Since we all agree for the need of ML now lets start off with Machine Learning.
What is ML according to me??

Although it is a vague version of what an ML is, The black box can be considered as our brain where we need to give inputs(images) of orange, apple, cat, dog, mop etc. and also tell the output of those images and train it.
Just like how a child learns by seeing pictures where we teach the child what an object is rather than telling properties right away. When an entirely different image is given which was not present in training input it should still identify it accordingly.
In a case of Dog vs Mop, identification considers 10,000 images of each is given to it and shown as dog and mop respectively. Now after it is trained if a new image is given which is completely not present in the 10,000 training images it must still identify it properly.
Here the answer(output) is shown for each image(input) but when a new image is given it must still identify it properly as a human being.
Children also after seeing multiple times images of dogs when they see a new breed of dog still identify it as a dog that is the whole concept here.
As seen above the property to identify each image is done by the algorithm itself rather than manually programming it. “The Machine is learning by itself” here where we just provide data for training.
So interested in Machine Learning yet?? understood why it is called Machine Learning??


We could see here that this model has failed in two cases out of 8, but let us cut it some slack it is even tough for us to determine them so it did a pretty good job out there identifying don’t you think?
Credits: Google Keynote for AI.
Next I have Machine Learning implementing example in 5 minutes coming up make sure to follow me on medium, linkedin, twitter, Instagram to get more updates.
Join our WhatsApp community here.
