But I try.
This experiment from Google turned me on to just doodling at bit.
Called Quick, Draw!, the application gives you something to draw.
You have 20 seconds to do your best.
As you draw, the application uses machine learning to try and guess what it is that you’re drawing. Now, the skeptic will say “that’s pretty easy; it told you what to draw”. If that was it, it wouldn’t be fun.
But it is fun; as you draw, the application “thinks” and when it recognizes something or some pattern, it will give you a hint as to what it thinks you’re drawing. When the application has it figured out, you’ll get an “Oh, I know…” and shoots you the answer.
You guessed it! It’s addictive.
When I first played around with this, I was on a laptop and used my trackpad to draw. This generated some pretty bad trials which improved a bit when I used my mouse to do the drawing. I’m sure that you know that my first attempts were pretty straight and not terribly well done. I graduated to my Bamboo tablet and then to the iPad. With a better device, I was able to do better drawings. You’ll notice that I never said that I ended up with something good.
But I tried.
The application claims that it learns based upon what you draw and that it uses the same sort of technology that Google Translate does. Over time, it claims that your drawings will make it smarter. That’s the key to successful Machine Learning.
I just hope my attempts don’t dumb it down!
It’s fascinating to watch the learning as you draw. There are definite Machine Learning “Ah hah!” moments when the application gets it. One object that I had to draw was a crown which ended up looking more like a baseball glove. But, it was drawing the points that gave it away to Quick Draw!
In the meantime, let your inner artist or your students’ inner artists put it to the test. It could lead to some interesting discussions about how “smart” computers are and can be – very appropriate for a discussion during Computer Science Education Week.