Artificial Intelligence and Machine Learning

Artificial Intelligence is a buzzword for many of us, especially around people who work on enterprise applications, legacy systems. With this article, I have tried to put together a quick read to enable us to understand AI and Machine Learning in the simplest possible way.

There are many definitions of AI but in the simple term “it’s a branch of computer science which enable machines to behave and work like humans”. AI was first introduced in the 1950s and it covers machine learning, deep learning, natural language processing, and expert system.

Having been introduced in the 1950’s it is only the last decade when AI has taken a big leap in the technology spectrum and the question comes in: why this area of computer science have taken a pace now and not before?

There are some reasons behind it and an important one is DATA, we are generating data at an immeasurable pace, which is a driver for AI. Now a days we have very high computational power, GPUs (graphical processing units) in place of CPUs, which provide us computational power to process images, videos at lightning speed. The third aspect is improved quality of algorithms, these algorithms provide better accuracy for predictive analysis. The last and most important aspect is funding, nowadays all big organizations are investing in AI, ML. Due to all of these factors development around AI has taken a significant pace.

Although many significant implementations are happening in AI and there are many which are being used in our daily lives, some of these regular ones are:

• Amazon prime’s or Netflix’s personalized recommendations by using viewers historical data, geography
• Alexa’s way of operating as per user instructions by using voice recognition technics
• Gmail SPAM filtering to put emails in the spam folder by using some specific words like lottery etc.

All of the above falls in the category of “AI narrow intelligence” in other words all of the above are under the weak AI category. The next stages of AI will be “AI general Intelligence” and “AI superintelligence“. Currently, machines are not capable of thinking hence we are almost stuck on narrow AI, and It is not predictable when we will reach the next stages of AI.

Insight in ML -?

There are many sub-categories within AI like machine learning, deep learning, neural network, image processing, voice recognition, and many more. In this article, I have covered machine learning briefly.

Machine Learning is a technology “which enables the machine to learn on its own by providing significant data as input, and improve further. And all these without being explicitly programmed to do the job”.?

Machine Learning vs Traditional Programming

To understand how ML works, I have compared it with traditional programming. I went through some of the material and concluded that the way machine learning works, it’s the opposite of the traditional programming approach, now you might think why it is so…

So let’s take an example of traditional programming

In traditional programming usual inputs are conditions/rules along with data and it gives us output/results.?

Example - So if we want to find the difference between table tennis racket and lawn tennis racket. In traditional programming, we will write the code and then supply different parameters like length, handle, base, size, grip, weight to check the results. If supplied data falls under those rules then the program will tell us whether it’s an image of a table tennis racket or lawn tennis racket. But it will need us to write quite a lengthy code to look at all different parameters.

In Machine learning, it works the other way around, and we swap Rules with Answers. In machine learning, we provide loads of correct answers (data) so our machine can learn the pattern and give predictions. For the above examples, we will provide loads of images of table tennis rackets and lawn tennis rackets, then the machine will learn the pattern by using ML algorithms and give Rules as output. This is a training phase of the model.

Let’s get further insight into ML, as we know Machine learning is a process that enables machines to learn and build a predictive model that can be used to find a solution for a problem statement. This is achieved by using the following main components:

• End Goal or objective
• Data (input)
• Model (based on algorithms)
• Predictions (output)

These components are an integral part of the following main stages of machine learning

? To make any ML model work properly we need to collate as much as data required, that process is data collection, nowadays many sources are available to provide you with required data. Collected data then needs to be cleansed to get into a usable format. Data is then divided into three parts - Training data, Testing Data, and Validation Data. Usually, 80% of data is used as training data, 10% is used as testing data and 10% is used for validation data. Training data helps the model to learn the right pattern, testing data is used to test the model, and validation data is used for model selection and fine-tune the model.

? In the next step we train machine by supplying a lot of training data to predefined algorithms, algorithms are a set of rules and statistical technique, which is used to learn patterns from data.

? Above step enables us to create the right model, which is the main component, and uses algorithms to predict the result.

? When the Model is created, we supply testing data to check if predicated results are giving correct results or not.

When the model provides correct results, it can be used for broader work or in production.

Artificial Intelligence and Machine learning both need significant learning and knowledge to become an expert but I hope this article will help you with a bit of insight.

Hope you enjoyed the quick read. Please provide feedback suggestions to improve it further.