Nowadays Talks: Machine Learning… What is that exactly?
As we all (some) know, early this year, Google-claimed Artificially Intelligence-based robot (which was bought by Google from DeepMind Company), AlphaGo defeated Go World Champion, Lee Sedol. What made AlphaGo different with its former AI-robot families was that DeepMind Co. made AlphaGo to be able to teach itself after being exposed with new patterns. That, is what Machine Learning term comes from.
Machine Learning is basically two words combined into one, literally. It means to let a machine (computer program to be exact) teach itself whenever new inputs are given. Prior to Machine Learning approach, AI were made with simple structural deep and/or multi-layered graphs with lots of choices and fixed set of patterns. With Machine Learning, there were no need to have those kind of hardships with many codes have to be written manually, since the program will automatically make choices from its past experience and fit it with the new inputs as a new pattern.
As an easy example… Let’s take a daily lives scenario:
- Tom’s family has a refrigerator.
- The refrigerator doesn’t contain any eggs.
- Tom wants to eat omelette.
- Tom goes to market to buy eggs
The previous method of AI will resolves only in these four patterns. But if they were given a new statement, it will bugged itself and made *ERROR* split out everywhere. With Machine Learning, the possibilities of new statement to be able to inserted into the already exist statements are being created by the AI itself, of course, with or without human’s checking.
———————————————————————————————-
As for me, when thinking about AI that can study by itself, I always reminded of several movies such as Robocop, Terminator, Robots, Wall-E and Big Hero 6. So what I think will be interesting to be built with Machine Learning approach for Artificial Intelligence is a helper/assitance robot. With how Machine Learning works, helper robots can give a nearly perfect satisfaction as long as they helped the right amount of assistance to the users.
Their stock will be limited and can will be lent for a period of time at first. The purpose for that is to check, store, and manage the new input the robot had gotten during their borrowed time. Cleaning, cooking, sweeping, sewing, working, etc. (unsuited learning materials accidentally created will be deleted each time the maintenance came), with time, the robots will gain enough information to be able to do all of those and can be commercially sold for use.
Although of course, we should never give full access to the internet for the robots or else something catastrophic might happen. :p
Created by yours truly,
Dionisius Andrian – 1901476245
3rd Semester 2016/2017
School of Computer Science, Bina Nusantara University