Everything You Need To Know About Artificial Intelligence.
What is Artificial Intelligence?
- Artificial intelligence is a field of computer science that embeds human-like intelligence into computers, and it is developed to extend human capabilities and not to replace humans.
- AI enables computers to find the common pattern in data without humans having to code them manually.
- The only innate intelligence that computers have is what we provide to them. computers use that intelligence to learn from the examples and create machine learning models based on the inputs and desired outputs.
Types of Artificial Intelligence:
Applied AI is used to do only a specific task. Examples for this type of AI are language translators, self-driving cars, virtual assistants, recommendation engines, AI-powered web searches, and spam filters.
This type of AI can perform a wide variety of independent and irrelevant tasks.
This AI type is self-aware AI that it can completely imitate human-level intelligence.
It cannot learn new things and make decisions from training data.
It can learn new things from experience to solve problems and it does this by teaching itself new approaches.
But still, we can't figure out what kind of consciousness that we able to give to it. So, it is improbable to say that we will create conscious AI in the near future.
Task Domains of AI:
speech and voice recognition
Natural language processing
Reasoning and motion
Related Concepts and subsets of AI:
- It is a subset of artificial intelligence that provides computer the ability to learn, without being explicitly programmed.
- Machine learning models are the algorithms used to find common patterns in the data without the involvement of humans.
- It does not follow rules-based algorithms instead of that it develops models to classify and make predictions from data.
- In machine learning, we split datasets into three subsets. They are training, validation, and test sets.
- Training subsets are the datasets used for training of algorithms.
- Validation subset is used to validate the result and to tune the algorithm parameters.
- The testing subset is the dataset that never used before, and it is used to evaluate how good our model is.
Types of Machine Learning Techniques:
- It relies on datasets with a class label, and we use these datasets to build classification models.
- Human-labeled data is used to train the algorithms.
- The accuracy of the model depends on the volume of datasets, which have provided for the training. As the volume of training datasets increases, then the accuracy of the result provided by the model will also increase.
- The supervised learning subdivides into three categories, and they are regression, classification, and neural networks.
- The datasets used for unsupervised learning does not contain class labels and letting the algorithm to discover class labels from unstructured data.
- This type of learning is helpful in clustering the data, where the data grouped according to how similar it is to its neighbors and dissimilar to everything else.
- It uses a reward function to penalize bad actions or reward good functions.
- It has provided with a set of rules and constraints, and letting it learn how to achieve its goal.
- Deep learning layers algorithm to create layers of neural networks that imitate the structure and function of the human brain.
- It continuously learns from unstructured data to improve the accuracy of results.
- It enables the natural language understanding of AI systems and allows them to work on context and intent of what is being conveyed by a sentence.
- It does not directly connect input and output rather than that it contains several layers of processing units, which connects the one layer of output to the input of another layer.
- It contains many layers between input and output, and that's why it is called deep learning.
- The Engineers and developers decide the number of layers and the type of functions that connects the layers. Then they train the model with lots of examples.
Artificial Neural Networks:
- Artificial neural networks are developed to mimic biological neural networks.
- It is the collection of small units called neurons, which is a computational unit. It takes incoming data and learns to make decisions over time like the human brain.
- Neural networks learn through a technique called backpropagation. Backpropagation uses a set of training data that match known input to output.
- A collection of neurons is called a layer. A layer takes input and provides output. Each neural network contains one input layer and one output layer, between these input and output layers it contains many hidden layers.
- A neural network that contains more than one hidden layer is called a deep neural network.
Types of Neural Networks:
- This is an uncomplicated and old neural network.
- This is a single-layered neural network in which the input nodes connected directly to an output node.
- The input layer transfers the input values to the next layer by multiplying it with a weight and summing the result.
Convolutional Neural Network (CNN):
- CNN is a multilayer neural network that mimics the animal visual cortex.
- A convolution is a mathematical operation that applies one function as an input to other function and provides an output, which is the mixture of two functions.
- Convolution is good at detecting simple structures in an image and mixes those simple structures to form a complex feature.
- Image processing, video processing, and natural language processing are the main applications of CNN.
Recurrent Neural Networks:
- It is known as recurrent neural networks because it repeats the same task for every element of a sequence by feeding the prior outputs as an input to the subsequent stages.
Applications of Artificial Intelligence:
Artificial intelligence is a field which has shown discernible growth over the past decades. It has vast application in various fields. Here, I have mentioned some of its application.
Natural Language Processing:
- Natural language processing is the subset of artificial intelligence that enables computers to understand human's natural language with the help of machine learning and deep learning algorithms.
- It breaks down sentences grammatically, relationally, and structurally to discern the semantic meanings of words.
- It might understand the emotional intent of a sentence, which means it might understand whether you are asking a question out of frustration, depression or confusion.
- Computers convert text-to-speech and speech-to-text with the help of natural language processing, which enables computers to communicate more interactively with humans. Text-to-speech is also known as voice synthesis.
- In business, AI-powered voice synthesis is used to enhance the customer experience.
- In medicine, voice synthesis technology is used to help ALS patients to regain their true voice instead of using a computerized voice.
- Computer vision focuses on the replication of the human visual system, and it enables the computer to process the objects in the images and videos in the same way as humans do.
- This technology helps self-driving cars to make sense of their surroundings.
- It plays a crucial role in facial recognition applications in which computers match the facial images of the people to their identity.
- It plays a vital role in augmented and mixed reality in which computing devices are allowed to overlays or embed virtual images on the real world images.
- It is helping doctors to arrive with a preliminary diagnosis by finding symptoms in X-ray and MRI scans and it can even detect cancerous moles in skin images.
- Humanoid robots have become possible due to the advancement of artificial intelligence. These robots are designed to do specific tasks.
- Android robots are kind of humanoid robots that aesthetically resembles human. For instance, Ai-Da is an android robot created to do paintings.
- Collaborative robots(cobots) are used to lift heavy containers.
- Robots are used to trigger specific movements in the human body to create new neural pathways in the brain and it is done by detecting patterns in massive movement related datasets of patients with neurological damage.
© 2019 Kavitha A