Artificial Intelligence — A bird’s eye overview on the current trends in the field
Artificial Intelligence — A bird’s eye overview on the current trends in the field
Abstract :-
Artificial Intelligence in the simplest of terms can be understood as the effort on the part of humanity as a whole to recreate and simulate the different parts of the cognitive capabilities of the human brain in technological environments and ultimately enhance and exceed those capabilities.
The groundwork of this emergent and highly speculated upon field of computing was laid down by John McCarthy in at a workshop in Dartmouth College in 1956.
Ever since the field emerged, it has caused massive stirs in the higher whole of technology of and continues on an upward conquest. From creating buzzes in reality shows, to allowing for self driving cars, to running semi-automatic robots, working in healthcare and even running businesses, the potential of AI can be called a form of infinity.
Introduction :-
“Intelligence” is a word which is very heavily researched and talked about as far as sentient life-forms are concerned. This perhaps stems from the fact that humanity as a whole evolved to be the most dominant species on planet Earth by using intelligence as its most powerful asset. Thus it is no surprise that intelligence is used a means of measuring success and progress.
As the advent of computing technology started gaining pace in the twentieth century, the question of being able to replicate the cognitive features and abilities of the human mind and ultimately modify, amplify and surpass them came into existence. The groundwork of this field of research is what came to be known as “Artificial Intelligence”.
Machine Learning — The first pillar of AI :-
Machine learning, which is a subset of AI, basically deals with the analysis of data and making intelligent decisions based on the results. It uses a new form of algorithms in which models are created to classify and make predictions from data.
Traditional algorithms are mathematical entities which often use the if-then-else construct to determine the most efficient courses of action.
Now, such algorithms do have their uses but they have their limitations which can only be overcome by using the machine learning models.
Machine Learning provides three modes of learning :-
i> Supervised Learning — The algorithm is trained on human-labelled data. The more samples we provide, the better the learning becomes.
ii> Unsupervised Learning — Here, the data is provided but not the labels. The machine is tasked with labelling the data. It is highly useful for clustering data which is grouped in accordance with the similarity and dissimilarity with other data sets.
iii> Reinforcement Learning— Here, the machine is provided with the data along with some rules and restraints. The machine has to find the best solution. A game of chess can be played using such a process.
Deep Learning — The second Pillar of AI :-
In a way, deep learning is a specialized subset of Machine Learning. It actually layers algorithms to generate a neural network or a replication of the structure and various functionalities of the sentient human brain. It allows an AI system to learn continuously and increase its efficiency and accuracy with regards to its results round-the-clock.
This is what enables modern AIs to learn really well from unstructured data.
Deep Learning algorithms do not directly map inputs to outputs. Instead, they make use of multiple layers of processing units. Each of the layers passes its outputs to its subsequent layers.
The number of layers and the types of functions which connect the outputs of the layers with the inputs of the next layers is decided by the engineers and developers when creating and developing deep learning algorithms.
Deep learning fixes an old issue of Machine Learning. As the volume of data sets increases, the performances of Machine Learning algorithms decrease while Deep Learning Algorithms keep on seeing rises in their performances.
Deep Learning has several use cases including image captioning, voice recognition and transcription, facial recognition, medical imaging, and language translation and many more. Deep Learning is also one of the main components of driverless cars.
Neural Networks — The Third Pillar of AI
An artificial neural network can be thought of as a collection of artificial “neurons” which are designed based on the way the human brain processes information. They take incoming data, process the same and produce decisions, just like the human brain.
Neural networks learn through a process which is referred to as back-propagation. Basically, the known and desired outputs are matched to see how well the AI is producing the desired output. Initially, the input is plugged into the network to determine the AIs output and then an error function determines the difference(s) between the AIs output and the desired output.
A collection of neurons is referred to as a layer. A layer can take in inputs and exude outputs. Each and every neural network typically has one input layer and one output layer and it also has some hidden layers. Hidden layers take in a set of weighted inputs and produce an output through an activation function.
Neural networks with more than one hidden layers are referred to as deep neural networks.
Perceptrons are the oldest types of neural networks. They are single-layered neural networks consisting of input nodes connected directly to an output node. Input layers forward the input values to the next layer, by means of multiplying by a weight and summing the results.
Convolutional neural networks or CNNs are multilayer neural networks that take inspiration from the animal visual cortex. They have high usefulness in areas like image processing, video recognition and natural language processing. Now, we may understand a convolution as a mathematical operation where a function is applied on another function and the end result is a mixture of the two. Convolutions are quite adept at detecting simple structures in an image and putting these features together to create more detailed features.
Recurrent Neural Networks (RNNs) are so called because they repeat the same process with all the elements of a sequence with prior outputs feeding the subsequent input stages.
Natural Language Processing along with Speech Synthesis — The Fourth Pillar of AI :-
Natural language processing is a subset in the field of Artificial Intelligence which allows an AI to process, understand and replicate the various human languages and speech patterns. Natural language processing uses machine learning and deep learning algorithms to discern a word’s semantic meaning.
What it does in this case is that it breaks down the sentences into segments ; grammatically , relationally , structurally and tries to grasp the context of the usage of the words. It can discern whether the word “cloud” refers to the natural clouds or cloud-technology. It can also discern and find out emotions from the writings or voices of different entities.
Natural Language Processing is what allows for the conversion of text-to-speech and speech-to-text.
Computer Vision — The Fifth Pillar of AI
Computer vision is that subset of AI which deals with the interpretation, understanding and implementation of the visual features of the human brain.
Using digital images from videos and other devices and deep learning models, machines can accurately identify and classify objects — and then react to what they — in simple words — “see”. It allows the digital world to interact with the real world.
Self-driving cars use computer vision to interpret and interact with their surroundings. Facial recognition systems also rely heavily on Computer Vision.
Two more important applications of AI:-
i> Robotics and Automations : The most commonly thought-of case whenever AI is mentioned is that of the AI driven automations. This has been made possible in several countries like Japan , USA , China and many more. AI technology is used to make robots do some preliminary works. Certain semi-sentient automations have also been seen. Most notably the Sophia AI , which has been granted Saudi Arabian Citizenship demonstrates the capabilities of AIs.
ii> Healthcare : AI has the potential to change the very way in which the field of healthcare is perceived. Preliminary diagnostics and even un-manned surgeries can be made possible by AIs.
Concerns and Ethical Issues Regarding AIs :-
As the field of AI continues to grow and emerge certain concerns and ethical issues have started to develop around the field.
The most famous issue is the topic of a fully-sentient unshackled AI.
People fear that at some point of time AIs would exceed human beings and revolt against them, ensuring widespread destruction.
Racism is also an issue as improper training of AI can lead to some racist outputs, like relating kitchen and laundry with women and shootouts and sports with men. Certain skin colours may get preferential treatment.
Privacy is also an issue as AIs might spill out consumer details unless security is taken into account.
Accidents can also be caused by AIs if their decision making capabilities are not worked upon. For example, we can think of collisions caused by self-driving cars.
Conclusion :-
“Artificial Intelligence“ is really a very interesting topic to work upon. The implications which it suggests lead us to ponder over the colossal way in which the Future could be enhanced or destroyed by this technology. It is safe to assume that in the near future AI will embed itself with all other forms of technology and may as well become inseparable from human lives. Thus , I will conclude by saying that it is up to us to use this technology wisely and thoughtfully.
