It’s very common these days to come across these terms – data science, artificial intelligence, machine learning, deep learning, neural networks, and much more. But what do these buzzwords actually mean? And why should you care about one or the other? I’m trying to answer these questions in this post, to the best of my capacity. But then again, I’m no expert here. This is the knowledge I’ve gained in the last few years of my data science and machine learning journey. I’m sure most of you will have better and easier ways of explaining things than I do, so I’ll be looking forward to reading your comments down below. Let’s get started then.
Data science is all about data, and I’m pretty sure you already knew that. But did you know that we use data science to make business decisions? I’m pretty sure you knew that as well. So what else is new here? Well, do you know how data science is used to make business decisions? No? Let’s look at that then.
We all know that every single tech company out there is collecting huge amounts of data. And data is revenue. Why is that? That’s because of data science. The more data you have, the more business insights you can generate. Using data science, you can uncover patterns in data that you didn’t even know existed. For example, you can discover that some guy who went to New York City for a vacation is most likely to splurge on a luxury trip to Venice in the next three weeks. That’s an example that I just made up, might not be true in the real world. But if you’re a company offering luxury tours to exotic destinations, you might be interested in getting this guy’s contact number.
Data science is being used extensively in such scenarios. Companies are using data science to build recommendation engines, and predicting user behaviour, and much more. All of this is only possible when you have enough amount of data so that various algorithms could be applied on that data to give you more accurate results.
There is also something called as prescriptive analytics in data science, which does pretty much the same predictions that we talked about in the rich tourist example above. But as an added benefit, prescriptive analytics will also tell you what kind of luxury tours to Venice a person might be interested in. For example, one person might want to fly first class but would be fine with a three star accommodation, whereas another person could be ready to fly economy but definitely needs the most luxurious stay and cultural experience. So even though both these people will be your rich clients, both of them have different requirements. So you can use prescriptive analytics for this.
You might be wondering, hey, that sounds a lot like artificial intelligence. And you’re not entirely wrong, actually. Because running these machine learning algorithms on huge datasets is again a part of data science. Machine learning is used in data science to make predictions and also to discover patterns in the data. This again sounds like we’re adding intelligence to our system. That must be artificial intelligence. Right? Let’s see.
Artificial intelligence, or AI for short, has been around since the mid 1950s. It’s not necessarily new. But it became super popular recently because of the advancements in processing capabilities. Back in the 1900s, there just wasn’t the necessary computing power to realise AI. Today, we have some of the fastest computers the world has ever seen. And the algorithm implementations have improved so much that we can run them on commodity hardware, even your laptop or smartphone that you’re using to read this right now. And given the seemingly endless possibilities of AI, everybody wants a piece of it.
But what exactly is artificial intelligence? Artificial intelligence is the ability that can be imparted to computers which enables these machines to understand data, learn from the data, and make decisions based on patterns hidden in the data, or inferences that could otherwise be very difficult (to almost impossible) for humans to make manually. AI also enables machines to adjust their “knowledge” based on new inputs that were not part of the data used for training these machines.
Another way of defining AI is that it’s a collection of mathematical algorithms that make computers understand relationships between different types and pieces of data such that this knowledge of connections could be utilised to come to conclusions or make decisions that could be accurate to a very high degree.
But there’s one thing you need to make sure, that you have enough data for AI to learn from. If you have a very small data lake that you’re using to train your AI model, the accuracy of the prediction or decision could be low. So more the data, better is the training of the AI model, and more accurate will be the outcome. Depending on the size of your training data, you can choose various algorithms for your model. This is where machine learning and deep learning start to show up.
In the early days of AI, neural networks were all the rage. There were multiple groups of people across the globe working on bettering their neural networks. But as I mentioned earlier in the post, the limitations of the computing hardware kind of hindered the advancement of AI. But from the late 1980s all the way up to the 2010s, machine learning it was. Every major tech company was investing heavily in machine learning. Companies such as Google, Amazon, IBM, Facebook, etc. were virtually dragging AI and ML PhD. people straight from universities. But these days, even machine learning has taken a back seat. It’s all about deep learning now. There’s definitely been an evolution of AI in the last few decades, and it’s getting better with every passing year. You can visualise this evolution from the image below.
Let’s talk about machine learning now. Machine Learning (ML) is considered a sub-set of AI. You can even say that ML is an implementation of AI. So whenever you think AI, you can think of applying ML there. As the name makes it pretty clear, ML is used in situations where we want the machine to learn from the huge amounts of data we give it, and then apply that knowledge on new pieces of data that streams into the system. But how does a machine learn, you might ask.
There are different ways of making a machine learn. Different methods of machine learning are supervised learning, non-supervised learning, semi-supervised learning, and reinforced machine learning. In some of these methods, a user tells the machine what are the features or independent variables (input) and which is the dependent variable (output). So the machine learns the relationship between the independent and dependent variables present in the data that is provided to the machine. This data which is provided is called the training set. And once the learning phase or the training is complete, the machine, or the ML model, is tested on a piece of data which the model has not encountered before. This new dataset is called the test dataset. There are different ways in which you can split your existing dataset between the training and the test dataset. Once the model is mature enough to give reliable and high accuracy results, the model will be deployed to a production setup where it will be used against absolutely new datasets for problems such as predictions or classification.
There are various algorithms in ML which could be used for prediction problems, classification problems, regression problems, and more. You might have heard of algorithms such as simple linear regression, polynomial regression, support vector regression, decision tree regression, random forest regression, K-nearest neighbours, and the like. These are some of the common regression and clustering algorithms used in ML. There are many more as well. And there are a lot of data preparation or pre-processing steps you need to take care of even before training your model. But ML libraries such as SciKit Learn have evolved so much that even an app developer without any background in mathematics or statistics, or even a formal AI education, can start using these libraries to build, train, test, deploy, and use ML models in the real world. But it always helps to know how these algorithms work, so that you can make informed decisions when you are to select an algorithm for your problem statement. With this knowledge of ML, let’s talk a bit about deep learning now.
Deep Learning (DL) is an advancement of ML. Even though ML is super powerful for most applications, there are situations where ML leaves a lot to be desired. That is where deep learning steps in. It is generally believed that if your training dataset is relatively small, you go with ML. But if you have huge amounts of data on which you can train a model, and if the data has too many features, and if accuracy is super important (accuracy is always important though), you take the deep learning route.
It is also important to note that deep learning requires much powerful hardware to run on (mostly GPUs are used), it takes significantly more time to train your models, and it is generally more difficult to implement compared to ML. But these are some of the compromises that you have to live with when the problem you’re trying to solve is that much more complex.
You might have heard of TensorFlow, which is a neural network that Google is extensively using and pushing to developers. Well, that’s using deep learning, as neural network is a kind of deep learning model. The self driving cars we started seeing in the last few years, they are self driving thanks to deep learning. There are many such applications of deep learning in the modern world that are kind of behind the scenes. For example, entertainment services such as Netflix are using deep learning extensively to improve their recommendations for you, and also to decide, based on user engagement, which shows are worth continuing production, and which shows need to be axed because they’re wasting time and money.
Most virtual assistants we see today, such as Alexa and Google Assistant, use deep learning to understand the requests you are making (Natural Language Processing — NLP), the tone, the emotion you’re displaying, and also to authenticate your voice in some cases. Fake news is a big deal today. Companies are being sued left and right for failing to control fake news propagation on their social platforms. So many such companies have started using deep learning to detect fake news items being circulated on their platforms and then take necessary actions. So yeah, deep learning is a big deal today.
There’s a lot more to write about all these different technologies that I mentioned in this post. These subjects are so vast that thousands of people have dedicated their lives to researching and bettering these technologies for the betterment of humanity. There’s also another group of people which believes that advancing AI is pushing humanity to extinction. But then again, none of us have seen the future, and time travel is still not a thing. Maybe, someday in the future, there’ll be an AI powerful and intelligent enough to come up with the design for a time travelling machine.
But now that we’re here in the present, and experiencing the revolutionary evolution of AI first hand, I feel we should embrace this brave new technology and figure out ways in which we can utilise it to make the world a better place. There are so many organisations and private firms researching on how we can use AI in healthcare to detect diseases early on and prevent lose of lives. There’s a lot of research going on to figure out how AI can help in cancer treatment. People are trying to discover new drugs which could open up a whole new window of opportunities for treatment. Things like medical imaging is also a field in which AI is being used to a great extent. Because people are using AI with GPU cores (deep learning) for medical imaging, and because the analysis of the images is also done using AI, we are seeing some great progress in the process of early detection of illness, accurate detection of illness, and timely measures to avoid life threatening diseases.
The financial industry is using AI heavily to detect frauds in financial transactions. With Keras and TensorFlow growing on a daily basis, we’re see new capabilities in fraud detection and prevention. Banking and financial institutions are able to study transactions, banking history, and credit scores of millions of people to detect and prevent loan and insurance frauds from happening. This has been crucial in saving billions of dollars in the last few years.
The luxurious future of being chauffeured by a robot is becoming a reality, thanks to AI and innovations in self driving car technology. I’ll also go on to say that in the future we’ll have small devices the size of modern smartphones that you keep at home to monitor various parameters of your health and body. We already have blood glucose and blood pressure monitors like that, and the latest Apple Watch even has an ECG reader. So yeah, with such advancements in technology and smart devices, once you start giving them enough data about your health, these devices and services will able to uncover patterns in your health and predict the future, which will in turn help you make course corrections early on.
Do you feel you have a clear picture of the differences between these various technologies? I tried to give a few examples of various applications of these technologies as well. Hope it helped. If you are planning on taking this knowledge further and try out machine learning or deep learning yourself, I’ve curated a list of top five ML and deep learning courses on Udemy. Maybe you can enrol and become a data scientist in the future.