I used to think artificial intelligence as a lot of
if operators, I never deep learn (yes, its a reference) about AI until just recently. I know that AI is mostly used for classifications, and then do action based on that classifications.
Long time ago, before the Night King destroyed the wall and say “AI is coming”, I was in this project used to classify an apple, and measure the quality of an apple. So basically I got a bunch of excel documents to be processed, so I used C# to process it and structure the data so I can classify it later.
I never learned about those fancy stuff such as Bayes, KNN, neural network, etc. So I wrote my own classifier, I take few parameters from the excel data, and measure it.
The parameters were: color, acid level, and taste. Note that I do not know anything about apple or how they got this parameters from, all I know is that those parameters are used to classify the apple and rank its quality.
The data itself was actually already classified, I mean all the rows of color, acid level, and taste already got label on it. My goal was not to classify the apple, but to classify on which rank is the apple based on its quality.
So we discussed on the value of the color, because acid level and taste were already numerical value. We decided that green was less appetizing than red, so we give it less score, and then some shades of colors were also assigned its own score.
So my classifier processed the color, acid level, and taste into a normalized value (ranging from 0–1). From that normalized value, we know that apple with the highest value is the best. We classify the range into something like: 0–0.2 is bad, 0.2–0.4 is OK, 0.4–0.6 is good, 0.6–1 is nice.
Well my job was to do that classifier only, but I think from the normalized value they compared it with people favorites apple in certain regions and they can distribute the apple better and profits!
Recently I watched an amazing implementation of AI that makes me interested even more. It was a DoTA 2 match of professional players VS an AI. It was amazing and it was confusing for me. As I said, I used to think about AI being a lot of
if, so I imagined, how did they do this with
if? Do they teach it to read the game internal variables, but then what’s so difference about it compared to the normal DoTA bot? They can just do a lot of
if, like if the enemy is here, do this, if enemy health is around 50%, do this, but do they?
Well the bot was actually a combination of supervised and unsupervised learning, I mean what? Unsupervised? Supervised? What are those? But how? I thought AI was just about
ifs. This is when I started to type into google,
Apparently, I got a hit on Udacity course about AI, and thinking of taking the course as soon as I finished my Udacity’s Android scholarship program. As for now, I just watch YouTube videos on AI before I went to bed.
At least I wrote a classifier before, so I have a rough idea about it, and this is I know that I learned AI before I even know about it, and now that I knew, I will learn more about it.
What do you think about AI? Feel free to share and maybe give me some tips on learning about AI.