Artificial Intelligence

  • Artificial intelligence is considered as the engineering of creating machines and computer programs to perform tasks with a type of human intelligence and which are now reserved for humans.

  • Artificial intelligence makes the machine capable to learn from its experience.

  • It enables the machine to verify the data, learn from the data, and give its predictions that are based on patterns hidden in the data, or inferences that might be very difficult for humans to predict manually.

  • At present, Artificial Intelligence tackles the following issues:

  • General intelligence

  • Knowledge representation

  • Learning, Motion, and Manipulation

  • Natural Language Processing(NLP)

  • Perception, Planning, Reasoning, Problem Solving, and Social Intelligence.

Artificial Intelligence technology

World of AI

Talking about AI industry, it is found as a fragmented space with many AI providers. So, it becomes necessary for us to differentiate. Many of the big AI companies are offering DIY development platforms or cloud services. 

Some big companies such as Microsoft, Google, and IBM Watson keeping an eye over the brilliancy of AI. Nowadays, it becomes more honorable than any other field in technology respect.

AI works with a pattern of the act and learns which means to say that the technologies make the machines capable to learn from the outcomes and then return the response with efficiency.

Generally, sites come up with the items that are based on the recent searches of users, hence enhances its productivity. 

A self-driving car is based on the algorithm to sense its environment and work accordingly, runs safely on a road. 

Even in day to day life, we use personal assistants like Siri in apple and Alexa features of amazon are also its examples.

Elon Musk

The future of humanity is going to bifurcate in two directions:

Either it's going to become multi-planetary, or it's going to remain confined to one planet, and eventually, there's going to be an extinction event.

ITS ORIGIN

It was the time after world war II, a mathematician again involved in changing history with a question in his mind "Can machines think?", popularly known as Alan Turing. His works lead to the successful establishment of a vision about AI. 

This AI was considered under computer science vision. Its aim was to answer Alan Turing's question and this concept of AI gave rise to many other concepts. Many debates and sessions were conducted on this topic only and yet it was not explainable at that time. After a long time, scientists, namely Russel and Norvig, came up with a definition "the study of agents that receives percepts from the environment and perform actions." They had gone through many experiments and finally, they concluded that Turing's test allows an agent to act rationally. After that, many other definitions were proposed but some of them were not considered universal. 

The first category of AI is termed as Narrow AI, also be considered as “Weak AI”, that can operate a single piece of work well with a lot of limitations and guidelines. It is designed to perform limited tasks. It usually focuses on a particular task that can be performed more effectively. It is inspired by human intelligence and is completely based on that. IBM’s Watson, Image or speech recognition software, etc. are some of the examples that are based on Narrow AI.

Another category of AI is termed as General AI, also be known as “Strong AI”, that can be observed in day-to-day life. This kind of intelligence can widely and frequently be seen in popular movies, such as “Alita-Battle Angel”, “Avengers: Age of Ultron”, “Terminator: Dark Fate” and many more. The artificial machines of this category are much more similar to human beings and can be significantly used to deal with many problems. This includes robotics as well.

AI can be broadly divided into two categories:

Neural Networks : 

1. Single-Layer Perceptron - The foundation structure of your average neural net. This type of architecture is limited to modeling only linear functions. Hence can only perform simple binary classifications.

2. Multi-Layer Perceptron - More commonly known as a fully connected network. This architecture can be useful to model any function, due to one or more hidden layers. It can be used for classification and regression.

3. Convolutional Neural Network - The convolution Neural Network is the go-to architecture for image classification/object detection. This is due to its ability to learn spatial features and generalize them to unseen data.

4. Recurrent Neural Network - Recurrent Neural Network works by feeding the current input along with the state of the previous time step. RNNs are powerful for recognizing patterns in sequences of data.

5. Auto Encoder - Auto Encoder learns the representation of an input image via a lower-dimensional version of the input data. This architecture can be used for de-noising images.

6. Generative Adversarial Networks - GANs synthesize new instances of the data that can pass for real data. There are two networks, a discriminator and a generator. The generator is synthesized new data while the discriminator tries to identify if that new data is fake.

 

                                                   
 

Let's get familiar with some AI concepts:

     Backpropagation

     Perceptron

     Multi-Layer Perceptron

     Convolutional Network

     

Hey!!, Are you done with the above theory?

If yes, then you know your basics are on the right track.

We will let you know more shortly...

  • CREATED BY ANMOL VARSHNEY & PALAK GUPTA