Major advancements that happened in the AI/ML industry in past two decades

            Suppose you’re in Chennai and going for an interview. Your Google Duplex booked a cab at 11:15 AM. A Self driving cab came to your hotel and dropped you at 11:45 AM at your interview building. Your interview started at 12:00 AM and you came out at 1:00 PM. At 1:15 PM you were taking snacks, and then ABP news sent you a notification “Dear Manglam, here are the news for today. Typically replies within a second.”

pic courtesy: internet

            It may seem a dream 30 years ago but it has become a reality now. A basic thought which totally changed the paradigm of developing smart computers. In recent 30 years, the field of Artificial Intelligence has emerged at a certain level that it is expected that 70% of enterprise will implement AI in next 12 months. But it is not happened in a day. There is a long story behind the development of AI. Now, we will discuss about the history of AI and it subset Machine Learning.

            Story starts from 1666 AD when a philosopher Leibniz proposed that a thought and logic of human mind is just a combination of simple concepts. A basic concept came out of this theory that if we are able to create a machine which knows the simple concepts of human understanding, then it will definitely solve extremely complex problems ‘1’. In 1943, two philosophers Warren S. MsCulloch and Walter Pitts presented a mathematical biophysics paper named “A logical calculus of ideas immanent in nervous activity”, where they discussed the working of neurons of human brain. It later became the inspiration of basic working of artificial neuron. 

            Later, many philosopher gave certain arguments on Artificial Intelligence but the most important contribution was of Alan Turing which changed the paradigm of understanding of machines. Before Imitation Game, we thank that machines can only follow the instructions. But Turing Machine changed the concept of thinking by stating that ‘Machines can think too.’ He wanted to build a computer that behaves like a human in a way that a suspicious judge cannot even differentiate between a human and a machine. Before making this machine, we must ask, how we think. We usually find the characteristics that help us to recognize things. We know what is square, which has four equal length sides, they are perpendicular to each other and make a closed curve. If we give instructions a machine to understand a square then it will definitely recognize the square by finding given specifications. That's why the paper presented by Alan Turing is the benchmark in the field of AI. 

            In 1958, John McCarthy developed a Language LISP which became the most popular language for AI research. The term Artificial Intelligence was firstly coined by John McCarthy too.

In the late 50s, Frank Rosenblatt made the single layer neural network which he named as perceptron. Neural Network was the simplified model of the working of brain. This concept became the striking idea for Geoffrey Hinton who is also known as the godfather of AI. In late 70s, AI was the jerk for industry. Only few programmers were working on it. It boomed after 2006 when there was a huge data on internet and needed to understand. Hinton made certain algorithms on neural networks which were the basic concept of today’s Google translator, speech recognition and language understanding. 

The idea behind AI was to create some machines who can work like human. The barrier was to understand the thought process of humans. It was not like kaato, gholo aur lagalo. Humans learn from their experience. Before, we were building computers who were working on particular codes. They were just following the given instructions. Now, we needed some algorithms so that machine can learn from their experience, like us. AI and ML helped to create such algorithms. 

Deep Blue: In 1997, IBM created a program Deep Blue which was working on the AI model. It used to find best and suitable moves to play chess. This was the first program that defeated World Champion Gerry Kasparov. The news created so much hype. The program was made that it used to learn by playing again and again, the human learning method. This was the time when people understood the power of AI.

In the year 2006, Geoffrey Hinton presented a paper named, “Learning Multiple Layers of Representation” which later became the approach for Deep Learning. This method was based on the multi-layer neural network that could work on a top down model and generate the outcome.

Driverless car: In 2009, Google started to make a driverless car which was working on Artificial Intelligence and Machine Learning algorithms. In 2014, it passed the USA driving test.

Natural Language Processing: On the internet we usually share the data through text or speech. Our languages are very ambiguous and have many nuanced and elaborative meanings. Sometimes it becomes very difficult for us to understand the meaning of a line. Is it showing respect or a satire? It seems very difficult for a machine to understand such language. Google translator works on the same. The first humanoid robot Sophia uses text to speech recognition method which is a part of AI. 

Use of Sanskrit in AI: We want to have a language for machine which must be unambiguous. If a language is ambiguous then it’ll have different meaning of a single sentence. Like we have different meanings of a poetry. Even a single couplet have thousands of meanings. In the last twenty years, we used to have an unambiguous language which is accessible for machines to process. In my knowledge, Sanskrit is a language which has only one meaning of a sentence. The meaning doesn’t change even if we shift the words. This type of language is very useful for artificial intelligence.

Big reason to worry: It should not be the part of given assignment but this is most important. We are making machines which can think. They are becoming smart. There is no doubt that they are more powerful. Then a question arises that aren’t we making machines which can destroy humans? We created them to help us but is there any guarantee that they will not destroy us. We are making machines who used to think like humans. Human is greedy. He can do anything for its purpose. We destroyed earth for our needs. How can we consider that machines working on the basis of our mind will not have such greediness and anger? 

Special thanks: Iresh Mishra