Data and Desire: The chase behind Machine Learning and Data Science
Who would have thought a few years back, that on hearing the term “computer science”, the first thing that would come to the mind would be machine learning? Who would have possibly imagined that ML jobs in Silicon Valley would sell as fast as freshly baked buns on a winter morning? No one, maybe not even Elon Musk…..
But, what caused this boom to occur over a period of a few years is quite interesting. Predictive analysis though efficient models, using neural networks to determine probabilities of events such as stock market forecasting, natural language processing (NLP) for text generation, conversational UI, and many more have provided the platform for the rise of Machine Learning through the ranks of computer science containing the likes of old warhorses like app dev, web dev, game dev, cybersecurity, etc. However, rising through the ranks does not necessarily mean that they are forgotten, in fact, the importance they have assumed has never been so much before…
But are we exaggerating things? Are we stressing too much on data and designing algorithms to manage data that we are overlooking other potential places of growth? Do the traditional languages stand a chance in front of Python and its monsters, namely Tensorflow, Pytorch, Keras, Pandas, etc? Well, the answer is yes and no.
Machine Learning, as per many, is an integral part of computer science and has been around for ages in the form of sorting algorithms, statistical models, etc. However, it got recognized when researchers realized that it could be used to predict real-time events accurately. But, as per Michael Hochster, from Stanford University, statisticians have been doing that since a long time. So, why has Machine Learning assumed so much importance recently? The answer to this, is the difference in predictions. Computer scientists use ML to model data, build neural nets and make accurate predictions. Statisticians, on the other hand, don’t give much importance to prediction. For them, building models is more about observing disparities in data, how data varies with conditions, how can the modelling structure be changed to interpret data more effectively. Computer scientists, on the other hand, tend to use models to predict events rather than worrying too much about theoretical circumstances, thereby spending more time on anticipating probabilities and developing methods to improve success chances and less time on model structuring. But, where are we going from here?
Recent research shows that the number of ML models far exceeds the amount of data and many models are now obsolete. Also, the general conscience is that with Deep Learning, traditional ML models don’t stand much ground in the long run. Even if this seems quite disappointing, the silver lining is that ML has already been integrated with other fields and is not going anywhere.
Machine Learning is proving to be a huge factor in game development as developers can know use ML models to make predictions about events in games and also model the game accordingly. Currently, most games use NPCs(non-playable-characters), so most of the times it becomes quite easy for the human player to predict the opponent’s movements. With machine-learning-based NPCs, the opponents become smarter with time and learn to make movements based on the human plays. Also, a new possibility is using NLP to talk in real-time with the fictional characters in the game, and making decisions on the basis of those conversations, thereby taking the gaming experience to a whole new level. The trio of NLP, AI and core ML has the capability to revolutionize the gaming industry but it is still in its initial stages.
Another field that has a huge potential for integrating ML with its original design concepts is cybersecurity. In today’s world, with increasing cyber attacks, a system which predicts the onset of malware and prevents it can prove to be a life-changer and efforts are being made to do exactly that. A software from Microsoft in 2018, called the Microsoft’s Windows defender anticipated a Trojan attack early by utilising various levels of ML in its design. Also, ML can automate several trivial tasks which are originally done by humans like solving basic consumer concerns, helping people set up anti-malware protection on their machines etc, this will allow humans to devote time to solving specialized problems.
To summarize, the hype around Machine Learning is good as it is providing the traditional fields of computer science an impetus to explore and develop without worrying about the consequences. However, we must keep in mind that ML without theoretical computer science, probability, statistics, and linear algebra is like ice-cream without a cone. You can eat it but it won’t have the flavor you desire. So, carry out research in ML, develop new models but always keep in mind that this wouldn’t have been possible without a strong theoretical comp sci base.
