The risk of learning

This is a story from the time when I was very young. We were back from a day long school picnic, and everyone was busy sharing the memories of the event with friends. No one was really in the mood of attending the classes. Our teacher, maybe, was feeling the same, or at least realized it.

“Let us start a new game today, everyone will share one learning from the activities of yesterday.”
It was an excellent way to start a new day. It was a day away from school, and learning was not expected. Even then many had something to share.

“Some ants have wings, and they can also fly.”
“Water on the ground disappears faster when there is a wind.”
“The stones heat up more than the grass in the sun.”
“Deep is scared of crows, and the girls are scared of earthworms!”
“I would like to go for a picnic every day.”

Even though everyone was exposed to the same environment, what they picked up as learning was entirely different, and sometimes even unexpected things.

Simulating human learning using machines is a tough challenge. Most of the software development is deterministic. The ‘Turing’ machine, which forms the basis of all algorithmic computations, is fundamentally based on deterministic logic. It is so deterministic that the algorithm to generate truly random numbers is almost impossible to write.

Since we are used to this, it is quite natural for the IT professionals to expect deterministic output from a machine learning project. However every machine learning project is an experiment. What you will get out of it is not known unless it is over. This will definitely create challenges in terms of meeting management expectations, isn’t it?.

Are you ready to take the risk of learning?