Teaching a Machine To Learn = Artificial Intelligence
The more I think about it, the more I feel that the future is in artificial intelligence. Even if not for professional reasons, I get the feel that may be, the machines with artificial brains might make our life better. There are so many things that Siri does that actually makes life simpler. That is an everyday AI that is already making life livelier. More importantly, she speaks with an Indian accent, which makes it all the more endearing. She also understands Indian accent which is like seeing penguins in the local Mysore zoo.
Recently, I wrote about isolating the source of the problem in a complicated system. This post follows from that, at least thematically. Here is what I am thinking. Suppose you have a complicated system, with about 24 individual components. Why 24? I don’t know. 24 chromosomes and such. Anyway, 24 components. Each component can communicate with some predefined signals about their status. This includes errors (something went wrong) and logs (something went right). These errors and logs will be used by the problem solver to isolate the problems, and then fix issues. Then, the system is fixed and it works again.
Now, this is where I want to talk about pattern matching or rather, identifying patterns and then matching them with existing collection of patterns. This is what we do in our everyday jobs anyway. For instance, if I hear a laptop making noise like a old mammoth, I know that the hard drive is failing or the fan is bust or both. I know this because of my years of playing around with hardware and helping people with their old 5 year laptops which they hold on to like it is a dear friend, or a child. People are like that, and when it finally comes to throwing it away (or recycle), they start crying and stuff. The point is, patterns is how a lot of problems get solved. In fact, even in the Sherlock Holmes stories, he would constantly rant about how he is better than others simply because he has an extensive knowledge of past crime, and nothing is actually new.
So yes, pattern identifying, collection and matching go hand in hand. We store this information in our brain and when something similar happens in the future, we get to the solution thing faster. I am no expert on artificial intelligence but if something (a machine, software, service or app) can do this even to a limited extent, that could be termed as artificial intelligence. An immediate proof are the various keyboard systems that power our smartphones. The keyboard system keeps track of all the words that are typed, and the longer you use that software, the better it gets at predicting what you might type next. Artificial Intelligence at work.
In fact, I sometimes feel that this is precisely why Facebook purchased Watsapp. If I were to build a conversational chat system, I will need to analyze billions of conversation exchanges. You see, most people don’t follow the standard grammar. A person in Mysore will converse differently with another person in Bangalore. The same person might converse differently with different people. If a company could analyze (you know, identify patterns) such conversations at a large scale (like billions of users) and imagine how good they could get in terms of artificial intelligence.
That brings me to another project that Facebook is already running, wit.ai. yeah, that is a lame domain, but the service itself is great. It allows you to harness the power of Facebook’s artificial intelligence efforts (I mean, I am sure they are keeping the good stuff to themselves but some artificial intelligence is better than no artificial intelligence) via apps and services. I have been planning to work on one it and build my own apps, once Project TD goes on auto pilot. Ah! Artificial Intelligence.
At this point, I remain optimistic about this AI stuff. I think it is the next step in tech evolution, and input devices. Of course, I am obsessed with input devices, but more importantly, it will allow a whole new generation of people to interact with machines in a way that they never could.
Here is one scenario that I hope will play out some day. Woman walks into an ATM. As she opens the door, her thumb prints are already analyzed at the door handle. The bank automatically knows that it is her. Still, to be extra careful, there is a pleasant greeting (in the language of the woman’s choice) to which the woman responds. Her voice is matched, and her multi factor authentication is complete. By now, the door has been closed and the ATM is already activated. She does not have to swipe her card or anything. She speaks into the ATM (perhaps a cool dude animation is running on the screen) about what she wants (which is usually money withdrawal). The machine does the processing, and places money on the delivery box. Then, she holds the door again, and the machine does its authentication and lets her out.
The future man! Exciting times lie ahead of us!
