Myths About Machine Learning

Co-author: Navin Saini

Today we hear about artificial intelligence and its component technologies such as machine learning, deep learning, natural language processing, and speech recognition in our day- to-day life. Artificial intelligence is used in various products and applications that we are familiar with – whether it is Google search engine, voice assistants such as Siri, Watson, or Alexa, autonomous vehicles, product recommendation on e-commerce platforms, movie recommendations on Netflix, fraud detection in banking and finance, cybersecurity, and facial recognition on Facebook. Artificial intelligence-based technologies are used in various industry verticals such as healthcare, automotive, business, and manufacturing etc. However, there are certain misconceptions about machine learning, which we would like to highlight.

1. Myth: Machine learning, artificial intelligence, and deep learning are the same

   We often use the terms machine learning, artificial intelligence, and deep learning interchangeably, however, in reality they are not the same. Machine learning is a subset of artificial intelligence. It uses various algorithms to extract patterns or to learn from the training data and build mathematical models to make predictions on new or unseen data. Whereas, artificial intelligence is a much broader term which includes robotics, computer vision, natural language processing, and machine learning. Deep learning is a subset of machine learning, a non-linear approach, which uses multiple layered neural networks to extract higher level features from raw input data. Deep learning mimics human brain in processing data for making decisions.

2. Myth: No human intervention required for machine learning

   It is a common misconception that machines learn the behavior of systems without any human intervention. The various machine learning algorithms are in fact written by humans in various languages. Programming is required to prepare the data and build a model using selective machine learning algorithm as per the business problem to make predictions on unseen data. Once the model is deployed for prediction, continuous monitoring of its prediction accuracy and decision to retrain the model on new data is required, to find new meaningful patterns, for making more accurate predictions.

3. Myth: It is fairly easy to build a machine learning model

   It is not that simple and easy to build a machine learning platform. It requires lot of expertise in machine learning algorithms and statistical techniques, hands-on experience in exploring the data, preparing the data for training and building the model using the relevant machine learning algorithms, and appropriate hyper-parameter tuning. This also requires expertise in handling big datasets, big data frameworks for model training, deployment and monitoring.

4. Myth: Any raw data can be used to build machine learning model There is a false belief that any raw data can be used for training a machine learning model. Data need to be properly structured with irrelevant data filtered out for building a good predictive generalized machine learning model. Quality of input training data is a paramount factor to understand the relationships/patterns in data to build an effective, high precision model.
5. Myth: Machine learning platform will replace human beings (loss of jobs)

   People are scared that machine learning can take over their jobs. But in actual, machine learning will take over the redundant, time consuming tasks. Humans can use their time on business decision making instead of doing time consuming repetitive tasks which can be done by machine learning. There will be new job opportunities created for data handling, building the model, and using the predictions to make better business decisions.

   People are scared that machine learning can take over their jobs. But in actual, machine learning will take over the redundant, time consuming tasks

By having a clear understanding of machine learning and its usefulness in various industry verticals, we can build predictive models to solve various business problems to make better decisions, improved productivity and efficient running of the system, which is not possible through conventional techniques.

Reference:

https://www.proschoolonline.com/blog/myths-about-machine-learning

https://medium.com/@pedromdd/ten-myths-about-machine-learning-d888b48334a3

https://ayehu.com/top-machine-learning-myths/