Data and AI Learning Group

Explaining Artificial intelligence to my 9-year-old

By Balagopal Thotakura posted Fri January 22, 2021 12:27 AM

  

Explaining Artificial intelligence to my 9-year-old

“Dad, what is Artificial intelligence?” My 9-year-old son asked me recently. Thanks to the online virtual school during these pandemic situations, he hears these catchy words from his elder sister’s classes. Before I could respond, he showed the Wikipedia definition ‘Artificial intelligence (AI), is intelligence demonstrated by machines, unlike the natural intelligence displayed by humans and animals, which involves consciousness and emotionality.’ His next question was – “How can machines think? Will machines control humans like they show in the sci-fi movies?”

Real world examples of AI

I showed him an example. Our refrigerator starts playing music, if its door is left ajar of some time. It is smart enough to know when to alert us of an open door. In our electric iron, we can set a knob for each fabric type. It then turns off the heating based on the that, to maintain the optimum temperature for that setting. And then we have the word editors or even the web-browsers, that can do automatic spell and grammar check of what you write.

While these were simple examples that he could easily co-relate, at the other extreme are powerful systems like IBM Watson, which won Jeopardy a decade back. And somewhere in the middle are these cars which can automatically adjust the steering, seat and mirror positions, based on who is driving the car.

How does it work?

It works the same way as humans learn something new. We teach a child a language through repetition. We feed information and expose her to new words over a period. And most importantly, we correct her when she goes wrong. How does a plumber know his stuff? He has been trained on it over a period so that he can do his job right. He was corrected by his trainer, when he had gone wrong.

Machine Learning is very similar. We feed information (curate content) to the machine and provide it feedback. The more information that we feed, the more the machines “are trained to find patterns and features in massive amounts of data in order to make decisions and predictions based on new data”. The system could be smart enough to provide answers to the questions asked, while also providing the % accuracy of the answer. This would not be possible unless we provide it feedback and tell it when it goes wrong.

The refrigerator starts a timer when the door opens and if the door is not closed in 30 seconds, it plays a music. If the door is closed before that time, the timer stops. Pretty simple, right? And some one has possibly fed all the words from the dictionary and all the possible phrase combinations to the word processor. So, as you type in the tool, it knows when you do a spell or grammar mistake, and it can suggest you with corrective options.

Adding IOT to the perspective

Taking it to the next step, comes the Internet Of Things. We heard about a car, that would point you to the nearby fuel station when it is about to get empty. Or, the washing-machine which automatically sends a message to the service center if there is any part malfunctioning.

All the machine needs is - to be connected to the internet and be able to send and receive information. Apart from showing the reading on the fuel gauge, the car needs send that information to the server over internet. The server would respond back with the information about the fuel stations nearby.

Conclusion

So, all that we need to do, is to teach the machine what to do. But that needs to be done in its language. Every machine could have a different way in which it should be trained. My son seems to have understood now. He said, “Obviously simple things can be thought easily, and complex things are tougher”.

And for “Will machines control humans” – well, I don’t know. Hopefully not be in my lifetime.

1 comment
12 views

Permalink

Comments

10 days ago

Nice article. Thank you.