Machine learning is changing the world. The development of this technology is leading us to the 4th industrial revolution where artificial intelligence is assisting humans in their daily tasks. Considering the advances of the Internet and how computers helped us to a scale, we can consider machine learning like the next step into improving services via automatization.
However, machine learning is based on statistical models, some of them are actually very complex and they require a deep mathematical knowledge to understand them. This becomes a problem went we need to justify why a machine learning algorithm is making decisions in a specific way. For years, some companies have avoided machine learning because they couldn’t understand the decisions and they found the models unreliable in this aspect.
Imagine the following situation: you have a smart car who is reading signals. The car can read absolutely every signal everyday but one day the car starts failing reading the signals. you may ask yourself what’s the reason, maybe the wind, maybe a failure in the reading, maybe the rain or humidity. whichever the reason is if the car is driving someone inside the life of this person, and the other people around the road, is at risk.
Even though we won’t be able to test every possible scenario (at least now), we need to test these specific scenarios to understand what the limits in our machine learning systems are, even if we are not able to know how the algorithm is working, testing will help us to make it more reliable, and the more this technology advance the better our machine learning systems will be.
So if you want to start testing machine learning try MLighter and tell us what else do you want to see in the tool.