In this topic, essentially we’re gonna talk about Artificial Intelligence, Machine Learning & Deep Learning. I’ll give couple of examples to demonstrate underlying innate differences and bust few myths & fiction around these buzzwords.
Also, I’ll briefly touch upon what it takes to make a career in this emerging technology.
Now, let me begin with the question, What is AI?
John McCarthy who happens to be a pioneer in the field says, ”AI is science & engineering of making intelligent machines.” Over the years, many other experts also defined AI to the best of their understanding.
Some of them are: A branch of computer science dealing with the simulation of intelligent behaviour in computers. A computer system able to perform tasks that normally require human intelligence, such as visual perception, speech recognition, decision-making, and translation between languages. To further demystify AI, let me tell you that AI can be a whole lot of if-then statements, or complex statistical models mapping raw data to particular categories. Here, key point to note is that these if-then statements are simply rules explicitly programmed by humans. Put together, these if-then statements are sometimes also known rule engines or expert systems. For example, these rules engines may mimic intelligence of an accountant with knowledge of the tax code, who takes information you feed it, runs the information through a set of static rules, and computes the amount of taxes you owe as a result.
Coming to Machine Learning, Machine learning is subset of AI, i.e. every machine learning algorithm & its application is AI, but not vice versa. For example, rules engines, expert systems – could all be described as AI, and none of them are machine learning. Machine learning algorithms are intelligent in deciphering the hidden rules, patterns or information from the raw data by itself. Here, there are no explicit hand made rules by the humans. The “learning” part of machine learning means that ML algorithms attempt to optimize along a certain dimension; i.e. they usually try to minimize error or maximize the likelihood of their predictions being true. They are, in short, an optimization algorithm. Algorithms self-adjust their parameters during training based on the nuances & subtleties of the data.
Spam email detection can be one of the easy example to understand Machine Learning. Researchers initially tried to identify key patterns in the emails which were considered as spam, such as the sender and recipient, the message ID, date and time of transmission, subject, number of words beginning with capital letters and several other email characteristics. Most spammers try to hide their identity by forging email headers or by relaying mail to hide the real source of the message, etc. Based on these patterns researchers developed classifier models to categorise the incoming mails as spam or not spam.
Arthur Samuel, one of the pioneers of machine learning, defined machine learning as a “field of study that gives computers the ability to learn without being explicitly programmed.” in 1959. That means, machine-learning programs have not been explicitly entered into a computer, like the if-then statements. Machine-learning programs, in a sense, adjust themselves in response to the data they’re exposed to. In Machine Learning, Neural Networks are one of the important algorithms which happened to be inspired by human brain. They keep on measuring the error and modifying their parameters until they can’t achieve any less error. Now, let’s talk about Deep Learning Deep learning is actually is a subset of machine learning or Artificial Neural Networks. Deep artificial neural networks have revolutionised the whole industry by setting new records in accuracy for many real life problems, such as Computer Vision, Speech recognition, Speech Synthesis, Language Understanding, Text Generation, etc. For example, deep learning is part of DeepMind’s well-known AlphaGo algorithm, which beat the former world champion Lee Sedol at Go in early 2016, and the current world champion Ke Jie in early 2017. The term “Deep” in Deep Learning refers to the number of layers in a neural network.
A shallow network has one hidden layer, and a deep network have multiple such hidden layers. Multiple hidden layers allow deep neural networks to learn features of the data in a hierarchical manner, because simple features recombine from one layer to the next, to form more complex features. For example in face detection, deep neural network learn features based on pixel intensity and in the following layer based on these features, learn and identify sharp edges in the images. In the subsequent layer, sharp edges are used as input features and facial properties such as, eyes, ears, etc. are learnt and at last based on these facial features, algorithm decides whether there exists a face in the image. However, developing Deep Neural Networks is computationally expensive business and it is one of the primary reason why specialised hardware i.e. GPUs, are being designed & developed to meet the demand of researchers so that they can process data to train deep-learning models quickly.
If we have to extrapolate on what Arthur Samuel said about machine learning, we can say that Deep learning is a field of study that gives computers the ability to learn without being explicitly programmed but it actually tends to result in higher accuracy, require more hardware or training time, and perform exceptionally well on machine perception tasks that involved unstructured data such as blobs of pixels or text. What’s in the future? In order to make career in AI, we need to understand different roles which are available in the industry. Primarily if I have to classify the roles, I’d broadly classify them into two categories, i.e. Analytics & Data Science Consulting Oriented Roles & AI Based Product Development Roles. Both the roles have certain overlap and key among those overlap is the perseverance and desired acumen to explore & understand the hidden insights in data.
Preferred choice of programming language is either R or Python. In my experience, R along with Visualisation tools such as Tableau, Qliksense, etc. are more inclined towards Data Science & Analytics consulting where as Python is preferred choice of researchers & developers in the advanced field of AI such as- Natural Language Processing, Speech Recognition, Computer Vision, Reinforcement learning, etc .