Вы находитесь на странице: 1из 1

AI and Its Usage

Artificial Intelligence in simple terms is the concept of making computers or machine to learn and think in
way like Humans do. How do the humans learn and think, is an open question? Researches shows that
humans learns by doing mistakes and doing a job for a long time means through experience. So, this is the
same concept which we try to implement for making a machine Artificial Intelligent.

Artificial Intelligence is vast filed and there are many braches of it e.g. Machine Learning, Deep Learning,
Speech Processing and Natural Language processing. Although, there are many filed but there are few
concepts which remains same for most of them e.g. all of these branches need large amount of data to
train a network, this data could be in the form of value, images or may be speech signals. Most of the
times, we perform following tasks using AI i.e. Prediction of some events, Analysis of a Big amount of data
and then finding out the patterns in that data for may be Business analytics related application, image
classification and detection related application, speech recognition application and finally may be for
recommender system for marketing products and services.

The major advancement in this field is currently due to the invention of Deep Neural Networks (DNN) or
Deep learning. Deep learning can be divided into two stages i.e. Training and Inference. In Training we use
a large amount of data and large computing power to train a DNN and after training when we get our
desired accuracy then we use that trained system for Inference stage where we only use that system for
may be prediction, classification and detection. So, having a large amount of data and computing power
is the requirement of implementing Deep learning at training stage. For inference stage we try to optimize
the network and try to fit that system in low cost hardware with less resource to cater the issue of scaling
up our solution.

In the field of speech processing we apply same approach of training a network and then using that
network for inference in real time. The difference is only that now we will use speech data. There are also
some classical way of implementing speech processing and recognition which use stochastic algorithms
but those are not very suitable for different changing situations. So, DNNs specifically RNNs are used for
the purpose of speech recognition and other classification tasks related to speech data.

In the field of Computer vision, the data changes from values to images and major tasks remains same i.e.
classification of objects in the image and detection of that object in that image. There are some
applications in which we use video data. Computer vision applications also find its application in Robotics
field in which we try to make artificial intelligent robots for many purposes may be daily routine task of
dusting and automation etc.

In the field of Natural language processing, we use a big chunk of data available online i.e. text data. We
know that most of our data which is digital and available online is text data. So we use that data for may
be hand writing recognition system or in recommender systems implemented in search engines like
google etc. There are also some application where we try to summarize a document using Deep learning.

Вам также может понравиться