Is AutoML going to replace Data Science jobs?
The answer is No…! Here’s why. Everybody has the same question. As a Data Scientist, you are responsible for performing the following steps in the Data Science process.
1. Understanding Business Problem
First, you need to have a business understanding and domain knowledge so that you can understand a given business problem and then propose an appropriate solution. Many times you don’t have to use machine learning.
I have seen so many data scientists in the industry, who for many problems don’t use Machine Learning at all. They use a heuristic-based approach or a simple analysis in excel or python, pandas and they come up with their insights so in such cases there is no ml and there is no AutoML
2. Data Collection, Data Cleaning and Data Exploration
Let’s say they decide to use Machine Learning, now you have to figure out how are you going to collect data.
Are you going to use some of the current data sources in your organization?, Are you going to use web scraping to get data from the internet? or Are you going to buy data from some third-party vendor? All of these decision-making processes are going to be done by data scientists and they help collect the data then comes data cleaning followed by exploration, where you clean null values, clean outliers, and further a lot of things then comes building a model.
Now in building a model you have to do Feature Engineering, Model Selection, Hyperparameter Tuning, and so on…
3. Building Model
Building Model includes:
- Feature Engineering
- Model Selection
- HyperParameter Tuning
Now, this is the part that can be automated by AutoML. So Model Building can be automated by AutoML tools.
Usually, you give the train data and they help you do Feature Engineering, Model Selection, Hyperparameter tuning, and so on…
4. Collecting Insights
So the AutoML tool actually helps data scientists make their process easier and then comes Deploying the Model and Collecting Insights from it.
You prepare a presentation and present it to your stakeholders things. Like that now see only Build Model section can be replaced by AutoML as of 2021 and there are so many other steps like understanding the business problem, data collection, cleaning exploration, collecting insights, all these things cannot be replaced by AutoML.
These are the hard problems you know like problem formulation, understanding what the business needs, and figuring out the right solution.
It is very hard for the computer to figure it out with Business problems. So the steps like Understanding Business Problem, Data Collection, Data cleaning, Data Exploration, etc cannot be replaced by AutoML completely, and hence the data science jobs will only prosper in the coming time.
Any company will hire you as a data scientist for the following three core skills
- Business Understanding
- Analytical Skills
- Communication Skills
Python, SQL, Machine Learning, all these are tool skills that are surrounded by the core skills and you use the tool and core skills to solve a given data science problem, and the core skills are something that is very hard to be automated by computers.
So if you are planning to pursue a career in data science, have no doubts. The future is very bright and you and will have lots and lots of career opportunities. So go ahead learn the skills and I wish you all the best.