Human and Machine Learning

Let us go through how human and machine learning happens with a simple tea shop example.

Customer goes to tea stall around 11:30 am outside of office campus and asks for a tea with low sugar. Shop keepers learns that this customer preference is "Tea with low sugar".

Next day same customer goes and Shop keeper servers tea with low sugar without even being asked from customer. This makes Customer more happy as shop keeper knows his preference. This works great with 100% accuracy. Machine learning with this test data also can perform with 98% accuracy.

Next day, the same customer goes for tea around 3:30 PM. He looked very tense and thinking of something. Shop keeper now asks"Sir,Sugar kammi ya?" (Meaning Do you prefer Tea with low sugar?). Customer replies now he wants Tea with normal sugar. Shop keeper now learns if the same customer comes in the afternoon, his preference is tea with normal sugar. This also works great in machine learning,but now the variance of result needs to be explained with more variables the machine learning model becomes complex and more and more becoming specific to that customer. The same may not be applied for another customer.

Next day, the same customer goes with his female colleague for tea around 11:30 am. Now shop keeper's human intelligence comes into play. He wants to play safe? Just asking "Sir Medium ya?" (Meaning Do you prefer team with Medium level of sugar?). Now the outcome also becomes complex. It is not just with or without sugar, another one added here "Medium sugar". Customer also preferred the same. This also adds up complexity and more and more specific. Training this level balancing generic nature of model needs to be achieved in machine learning.

The way data and human behaves may not yield always the same result. Machine learning also gives probability of result between 0 to 1. That's why so many events could not be predicted in advance. Have we predicted this much scale of corona outbreak initially?