AI Bias: What to do? (Part 1)

Artificial Intelligence is a technology which covers industries from almost every possible field. We can say that it is a tool which is really helpful to us in many ways.

Basically, Artificial Intelligence is a huge set comprising of Machine learning,Computer vision, Natural language processing, Deep learning to name a few. One might be intrigued when they hear the term “Artificial Intelligence”. As in how can Intelligence be artificial? Is it not natural? Yes of course it is natural for humans, but how about the machines being intelligent? It sounds so cool when we get the maximum out of machines if they are intelligent right? So how can the machines be intelligent? Well, to understand all this we should know what intelligence means. Intelligence is not the ability of scoring marks or excelling at something but it’s the ability to think and respond through actions,words etc. In other words we can say that it means to act and think humanly along with thinking and acting rationally .That’s what it means when we say all are intelligent. So now that we know what intelligence exactly is, let’s see how machines can be intelligent. There are examples like Sophia,a humanoid robot which was granted citizenship in Saudi Arabia looking at the way it behaved just like a human. There are many such humanoids and robots in the world today which are artificially intelligent. It’s the data which aids the process. Data to the robot is like mitochondria to the cell, is the powerhouse for their knowledge,their behaviour and their actions.
It's just analogous to bringing up a child. If the child is trained well with good qualities ,he/she can be a good citizen. In one way we say that the surrounding should be good for children in order for them to be better citizens.There are many examples of children getting into wrong company and spoiling their lives. In our case here it’s the data which decides the future of the machine.A good unbiased data gives out a good result whereas the bad and biased data is a threat to the industry and the society,It can lead to many disasters. There are many movies which have shown this through their characters. Machines learn from the data which they are fed on.
Before training the machine, we need to know how the stage till training is reached which is when we understand where the data might go wrong.Data is a collection of different pieces of information from countless resources. More the data, better is the result. The data collected initially is very raw, just like the gold in a gold mine. Before making jewellery out of it, it undergoes several processes and during these processes the gold’s purity is determined based on carats. Similarly regarding data, this raw data collected from the resource undergoes determines the quality of the result we get finally.
The process of “data processing” or “Data mining”. Some major steps are:

• 1.Data collection: The first stage of data mining where the data is collected and stored.The data might come from any percepts like sensors,people etc.The data can be in any form Image,text or video.Later this data is stored in some suitable formats.
• 2.Data Treatment: This step mainly takes care of the data analysis for example the variables available,their range,the data types ,missing values are taken care.
• 3.Data transformation: The crucial step where the missing values are filled and some of miscellaneous data is dropped inorder to improve the quality of the data.
• 4.Data training:Before the data is being tested into real applications,It should be validated and tested well in order to perform well in any kind of variations.
• 5.Model development: Model refers to the the final file after the training.This can recognize the patterns.We use “Algorithms” to train the models.Algorithms are a set of instructions which are used on the model to help it identify the pattern.