Data science, machine learning and AI: do you know the difference?
29 January 2018 | Written by Filippo Scorza
I like to speak clearly so, let’s cut to the chace and let’s summarize the differences:
- Data science produces “insights” that allows understanding data
- Machine learning makes predictions
- Artificial intelligence produces actions
Wait, I would like to go beyond mere pure definitions and give you insights for each of these topics
No, nothing too difficult, I promise!
Data science stands out from Machine Learning (ML) because its goal is basically human: acquire knowledge and understanding.
Understanding can be divided into three perspectives:
- Descriptive: “the average customer of our e-commerce has a 70% probability of future purchase”
- Exploration: “customer B has a different service renewal rate than customer A”
- Random: “in a random test it has been shown that customer service B users renew the service more frequently than those of customer service A “
In Data Science it seems that the human being is always involved unlike the ML and the AI (not including the developers, obviously!).
So, Data Science produces graphs, analysis, linear regressions and embraces statistics having a considerable amount of data available.
That’s why we can say data are the new petroleum: because they allow us to understand and define the profiles of users, the market’s trend, and the economy.
Machine Learning provides predictions.
If you give an X instance with certain characteristics, machine learning can calculate the Y type predictions.
Ok, I know that’s not so clear.
Another example: if we provide machine learning all medical and health data and reports of 100.000 individuals with similar characteristics, machine learning will be able to predict, more or less accurately, if a new patient with similar characteristics can be subject to some form of the disease.
Or another one: if we give to a machine learning, all the computer chess games data, it will be able, not only to play but also, to make predictions about the possible future moves of the opponent.
We should imagine ML as a neutral algorithm that, thanks to autonomous or assisted learning, will learn everything.
Artificial intelligence produces actions.
AI is an autonomous agent but, in order to give a more clear definition, we have to compare it with machine learning.
For example, let’s suppose that a self-driving car should learn how to recognize the “stop” road sign.
To train the ML we have to build a dataset of millions of photos of roadside objects and we have to write an algorithm (imaging recognition) to predict which of them contain the stop signal. Once our car is able to recognize the stop signs, it will have to decide when to hit the brakes.
Timing, road conditions (dry or wet) and unexpected events are all part of the theory of control: and that’s why artificial intelligence can make a difference.
Basically, artificial intelligence will be able to make decisions autonomously.