Sunday, 25 May 2014

A Measure of Learning

How do you know if someone is learning?

For a long time, the mainstream opinion seems to be that you should have answers.
That is, if you know answers to certain questions in your field, you are considered to be learned in that field.

In machine learning, a similar conventional measure is used where the agent is supposed to produce the closest answers to specific problems.

But do we work this way?
"I think, therefore I am."

I have personally observed that people who ask the right questions are the ones who are really learning, especially amongst coders.

To build a measure that determines the ability of an agent to question stuff around it, and by extension, its own existence, could be seen as one that was truly measuring learning in the absolute sense - developing conscience.

I'm working on an algorithm along these lines.
Open to ideas.

Phew.

No comments:

Post a comment