AI is all around us. Be it the Face ID in the all new iPhone X or the self-driven cars. But we may not know much about this fancy sounding word. Let us break it down. AI simply refers to machines with brains. With this technology, they no longer need the human help in the form of commands or prompts. For example, Siri doesn’t require human assistance to talk to you. It just knows what to say. Not just that, machines with AI can also learn new stuff on their own as they encounter new situations or sets of data. Say, you were looking for a pair of shoes online the other day. Oddly enough, most of the ads then on in your apps are about shoes. Ever wonder why? It’s because someone, actually “something” did their homework.
Time to learn some more fancy words. AI is a very broad concept. It covers Machine Learning (ML) which again has a sub-branch called Deep Learning (DL). In a crude sense, ML is a student who makes notes but only to cram for exams, while DL reads a text, analyses it and then writes answers in her own words.
In ML, the machine doesn’t rely completely on the code of its creator. Computer code guides it only for the brief initial period to gather data and information. After that it is on its own. Say we want the machine to learn to recognize a horse. The coder would simply show pictures of several animals such as a zebra, donkey etc and point out the horse amongst them. After repeating this exercise a couple of times, it would figure out for itself how to distinguish the horse from the others. So basically, in ML, the machine ‘learns’ from the given data and prepares its own algorithms etc to extract important information. Based on this, it is also capable of making more profitable decisions. ML can be used to translate languages and recognize faces, voices and objects. The application of ML can therefore be seen in chatbots such as Siri and Cortana. If you remember, Google’s computer Deep Mind that won the Go World Championship, it was again ML.
DL is super cool if you ask me, it works almost like a human brain! Which by the way means that it is slightly complicated. ML can tell you it’s a horse, but it doesn’t know why. But DL takes it up a notch. The machine for starters uses ML and then creates artificial neural networks (ANN). Each neuron has an answer in the form of a yes or a no. There are layers and layers of such artificial neural networks hooked with one another. Thus by asking questions, it ‘learns’ new things such as what is what. Let’s take a very simple example. The machine is supposed to give you video recommendations on YouTube. Then a simple neural network would look like this:
Has the user seen any video uploaded by this channel? Yes (1).
Does the tag in the video say ‘comedy’? Yes (1).
Was it uploaded six months ago? No (0).
Is the content similar to those channels subscribed by the user? Yes (1).
So the code so formed is 1101 and bingo, it is thus eligible to be a video recommendation. However, DL is much more complicated. It is composed of millions of networks and deals with a huge amount of data.
These are the basics of AI. If you haven’t already, read more about Sophia, the humanoid robot. It is one of the finest breakthroughs achieved by AI. On the one hand, a lot of research is taking place in this direction, but on the other, the skeptical scientific community has raised a very ‘serious’ question – what if AI gets out of hand and the robots take over the world? So if you think about it, it’s a wonderful time to be alive!