Artificial Intelligence is consisting of two word “Artificial” and “Intelligence”. In simple word “it is a manmade thinking power”. Machine Learning is also consisting of two words, “Machine” and “Learning”. In simple word “A machine (Computer) capable of learning from past (data)”.
Artificial Intelligence ( Al) and Machine Learning (ML) are two most trending technologies of modem world. They are co-related to each other and capable of creating intelligent and smart system. Artificial Intelligence does not require to pre-program, but it works based on its own intelligence. It uses Machine Learning algorithms and Deep Learning neural network.
It is a new age tool of business management, defense, agriculture sector, weather forecast, health sector, financial management and disaster management. It is highly useful when data size is so huge that it is impossible to handle manually and draw a logical information. It also used for data mining from various sources (Digital, visual, oral etc.) and arranging the data for further processing.
Example- “Go straight and take left turn after 500 meters” is a well-known female voice, when “Google Map” is being used to reach any destination. Google Map guides and instruct the driver, suggest alternate route, estimate time of reaching destination, inform traffic and road condition etc. It is a live example of Artificial Intelligence being experienced and used by most of us on a day-to-day basis. Let us discuss Artificial Intelligence vs Human Intelligence.
A. Human Intelligence (Brain) and Artificial Intelligence
Advertisement
The human brain consists of two symmetrical parts with partition at center. The Left brain and the right brain. The Left brain controls the muscle of right side of body and vice versa. It is general believe that one part of brain is more dominant over other. Left brain dominant persons are more logical and methodical whereas right brain dominant personal are more emotional and creative (Out of box thinking, art etc. ). So, the human intelligence is classified under four categories.
- Think Logically
- Think emotionally
- Act logically
- Act emotionally
Artificial Intelligence are being developed to replicate above four intelligence of human being. Smart computers are capable to think and act logically, but not able to think and act emotionally. Artificial Intelligence imitate the logical or rational thinking of human being with ease at very faster speed.
B. Artificial Intelligence (Al)
It is a branch of computer science, which concern with building smart machine capable of imitating human intelligence. The basic purpose of Al is to perform certain task by substituting human intelligence. The ultimate goal of Al is to develop reasoning, learning and creating perception like human being by computers.
The growing size of data base (because of digitalization of information) and availability of powerful computer systems has facilitated, the faster growth in the field of Artificial Intelligence. Presently, it is virtually impacting the basic functioning of every industry and expected to dominate the future as well. It is being identified as most important tool for future generation. The industry, region or economy stronger in Artificial Intelligence will lead the world and create enough wealth for themselves.
It is neither advisable nor feasible to ignore development in the field of Al. Artificial Intelligence has a very wide scope i.e. is set of many subsets and most popular among all subsets are Machine learning and Deep learning.
Machine Learning – Under machine learning, past data are compiled and analyzed to formulate the behavior (mathematical equation) dependent variable vs change in independent variable. Thus, it is a projection of future event based on formulation of past data.
Deep Learning – Deep Learning is very closure to machine learning. Only difference is that the input data runs through biologically inspired neural network architect in Deep Learning. There is absence of various layer in Machine Learning.
The biologically inspired neural network is consist of number of layers and data is processed through them. The numbers of layers allow the machine to go “deep” in its learning for making connections and weighting input for the best results.
C. Type of Artificial Intelligence
The Artificial Intelligence are divided in three categories. Presently we are dealing with Weak Al and General Al. The Strong Al is the future of Al.
Narrow or Weak Artificial Intelligence (Weak Al) – It is a simulation of normal human mind and mainly focus on performing single task. It operates under various constrains and limitation with ease and high efficiency. It is mainly used for repetitive joh or routine activities.
Computer playing chess, auto driven car, heat seeking missiles, robotics used for painting or performing crucial job in Industries and detecting bank fraud by flagging unusual transaction (Deposit and Withdrawal) are good examples of Narrow or Weak Artificial Intelligence. With development of Technology the usage of Narrow Artificial Intelligence is increasing exponentially in Defense, banking, Health sector, manufacturing Industries, metrology department and space science.
Latest example of Weak Artificial Intelligence has been exhibited in Israel-Hamas Battle of May 2021. The Hamas attacked Israel with more than 1,000 missiles in a day. The Israeli “Iron Dom” identified and tracked the trajectory of all Rocket and Missile coming towards it and fired counter missile to neutralize incoming missile in air. The Iron Dome achieved the success rate more than 90% and saved Israel from major disaster successfully. The success rate and response time of Iron Dom was excellent and superior than human being.
al General Intelligence (AGI) – It is simulation of above average human mind. It performs many tasks with high accuracy under various constrains and limitations. If Weak Artificial Intelligence is equal to person of single skill, then multiskilled person is equal to Artificial General Intelligence.
Strong Artificial Intelligence (Strong Al) – Researchers are in process of developing Strong Al, which is equivalent to human intelligence. Artificial General Intelligence (AGI) augmented with Artificial Biological Intelligence (ABI) can replicate human intelligence. However, it is not yet developed and still in the mind of scientist and fiction writers.
D. Machine Learning
Machine Learning is part of Artificial Intelligence (Al). The basic steps of ML are collection of past data, analyze the behavioral pattern and make decision based on these patterns with minimum human intervention. Most important challenge of Machine learning is handling large data and finding right model of high confidence level. Machine Learning accepts structured and semi structured data.
Real-life data are collected from various source. So, the data formats are not uniform. It is available in various size / shape, messy and incomplete format. It is required to combine these data received from various source in different format into a uniform stream of information. So, pre-processing of data with help of specialize tools is most important steps of Machine Learning. Different type of data requires different type of module to process the data adequately.
Selecting best model is a time-consuming activity. Trial and error are the core of Machine Learning. If one algorithm fails, try the other one in a systematic approach, till most appropriate algorithm is located. The complete sequence of Machine Learning is as under.
- Data mining – Collecting data from various sources.
- Processing the data – Converting data of different size, shape and format to usable format.
- Derive algorithms – Derive best algorithms model based on data mined.
- Train module – Train the model to use the data successfully.
- Iterate – Iterate the model to identify the best suited module.
- Integration – Integrate the trained module to production system
E. Important Terminology
Important terminology used in Machine learning are as under
- Regression – It is a line or curve that passes through all data points in such a way that the vertical distance between graph line and actual data point are minimum.
- Dependent or Target Variable – The variable we wish to predict or understand.
- Independent Variable or Predictor – The variable which affect dependent variable.
- Outliner – Outliner are either very low value or very high value, which generally ignored while calculating or testing the module.
- Overfitting – When algorithms module works well with training data set but not with test data, it is called overfitting.
- Underfitting – When algorithm module neither work well with training data nor with test data, it is called Underfitting.
- Multicollinearity – When independent variables are correlated to each other more than dependent variables.
F. Type of Machine Learning
Machine Learning is broadly divided in three categories as under
- Supervised Learning – Supervised Learning builds models to predict outcome in case of uncertainty based on known set of input data and known set of response. The basic steps of Supervised Learning are
- Unsupervised Learning – It reveals hidden structures and pattern of a given data set. Clustering is the most common technique used in Unsupervised Learning. It is used for application like object recognition.
- Reinforcement Learning – It is mostly used in games.
G. Type of regressions
Various type of regression used in algorithms model, important among them are as under
Linear Regression: It is most simple form of regression and assumed to be of dependent variable of continuous nature. Model depicting annual rain fall and grain production. Grain production (dependent Variable) is dependent upon rainfall (independent Variable) in the area.
Model calculating arrival of sugarcane at sugar factory gate. Sugarcane harvesting quantity is dependent variable follow Linear regression models and depends upon no of harvester employed, which is an independent variable. The basic equation of Linear regression looks like
- Y= a+bX
- Y = dependent variables (target variables),
- X= Independent variables (predictor variables),
- a and b are the linear coefficients
Logistic Regression – The independent variable can be continuous or binary but dependent variable is always binary (Two categories). There is two outcome dependent variable (Win or Lose) in election dependence upon many independent variable like (Intensity of Election Campaign, Candidate profile, public perception etc.) The basic equation of logistic regression looks like
Polynomial Regression – It is technique to fit nonlinear equation by applying polynomial power of independent variable. The area of circle (dependent variable) is polynomial equation of radius of circle (independent variable) The Polynomial regression equation looks like photp, pg. 6
Ridge Regression – It is used in case of overfitting, when model perform well in training data but poorly in test data. Small amount of bias (penalty) is introduced to achieve long term better predictions. It is one of the most robust versions of linear regression.
Lasso Regression: It is another regularization technique like ridge regression except penalty terms contain absolute weight in lieu of square of weight.
Bayesian Linear Regression – It is a combination of Linear Regression and ridge regression. It is more stable than linear regression.
Support Vector Regression – It solve both linear and nonlinear models, whose algorithm works for continuous variable. The idea of SVR is to minimise error. Support vectors are the datapoints which are nearest to the hyperplane and opposite class.
Decision Tree Regression – It is efficient because of using strong algorithm for predictive analysis. It looks like a tree having root, stem, branches and leaves.
- Internal node represents the “test”
- Branch represent the result of the test
- Leaf node represents the final decision or result.
It is built with starting point, the root node/parent node (dataset ). The branches of tree splits into left and right as child nodes (subsets of dataset). The child nodes are further divided into their children node
Random Forest Regression – It performs regression as well as classification tasks. Most powerful supervised learning algorithms, which combines multiple decision trees and predicts the final output, (average of each tree output). The base model are combined decision trees. The basic equation is – h(x)= g0(x)+ gl(x)+ g2(x)+….
H. Arguments against Artificial Intelligence and Machine Learning and facts
Major arguments against Artificial Intelligence and related facts can be summaries are as under.
- Unemployment – Industries looking to automate certain activity with help of Al, will reduce manpower requirement and ultimately create mass unemployment. Specially for the country like India, where unemployment is quite high. Fact – Usage of Al will boost economic activity. Expanding economy will create many job opportunities. It will ultimately generate employment.
- Threat – Strong Al may take over human intelligence and human being may not be able to keep pace with computer. Fact – Al is supporting human activities and ultimately it is a tool of improving human efforts and not a threat. Al will complement the human effort of overall prosperity
- Usage of Al in the field of defense may create havoc and senseless destruction to humanity, if it falls in hands of ruthless dictator. Fact – These arguments are similar to the phrase “fear of unknown” or “resistance to change”, which is hypothetical in nature. Al will always work under the supervision of human being in supportive role and never as master of human being.
Conclusion
After considering various aspects of arguments against and in favor of Al, it is a big boon for technology savvy country like India. Artificial Intelligence is expected to boost India’s annual growth rate by 1.3% in 2035 (Niti Aayog, 2020). This is a clear indication of growth potential and future prospects of Al. The strong logical thinking of average Indian is having a natural edge over others in Al and ML.
Artificial intelligence is being successfully used for diagnosingpatient and treatingthemin case ofpandemic. Banking and Insurance sector are using for fraud detection. Defense is using to track enemy movement including enemy’s rockets and missile. Super markets are forecasting inventory requirement in normal condition and in case of natural calamities. Usage of Artificial Intelligence is expanding exponentially in 21st century and expected to continue in future too. It is expected that it may replacing the human intelligence in many applications.
Presently Al and ML is not having foot print in Finance and Accounting function except some specific function in Banking Industries. But Al and ML might be able to perform various accounting and audit task (Vouching, Budgeting, fraud detection, performance appraisal etc.) successfully in future. It is the new age tool having bright future prospects.