How it all started: AI, ML & DS!

Happy New Year 2024 to all! and Welcome to my first post this year and would like to present an introduction to AI, ML & DS.

I started with Data Science around 2019 and picked up a few basics (Hypothesis, Linear Algebra, Probability, Classification & Regression Techniques). With the help of R language did basic Data Analysis and simple ML Models. All these without even thinking about How this all started?, Why is it such useful in times to come?.

Modern-day Artificial Intelligence is somewhat a birthchild of great scientist "Alan Turing" whose initial question of Can Machines be intelligent? initiated a curiosity to design a Turing test. A Turing test is to imitate the human-like intelligence of a machine and make it indistinguishable from a human conversation and if humans cannot recognize if they are conversing with a machine then the machine passes the Turing test. The Turing test is a textual-based conversational approach and this has ignited the scientific community across the world. Improvements in that area by Academicians and Scientists gave rise to terming this human-like intelligence "Artificial Intelligence" by John McCarthy around 1956.

Humans have such remarkable intelligence why do we need AI?

This is the question that restricted me from starting on this subject quite late...The following reasons have made AI adoption faster across different sectors.

• Efficient ways of handling and processing big data and making quicker non-partial decisions consistently.

• Automating Repetitive tasks that consume human time which could be used for other creative and strategic aspects.

• Pattern Recognition at large scale, Simulation, and creating virtual worlds that can aid in analyzing situations and scenarios for different industrial and real-world purposes.

• Explorations at tough terrains like deep seas, mining operations, and other undiscovered places can be done through AI-controlled Robots

• Accuracy in identifying good/bad scenarios, validating persons in huge crowds or important public gatherings through Computer vision technologies.

• Medical diagnosis to predict future health issues, aids doctors in analyzing psychological issues, and also helps persons with disabilities to access real-life needs and communicate freely.

This is just a small list, and the possibilities are endless as wherever human intelligence could be applied AI can work along.

Ok, so where does Machine Learning fit in?

Machine Learning (ML) had its inception in 1959 when Arthur Samuels (of IBM) was fascinated by the idea of machines that could mimic human learning and decision-making. That enabled him to coin the term "Machine Learning". It is quite logical how the Interest and work in Artificial Intelligence in the early 1950s has given rise to the Machine Learning and Data Science fields.

Simply put, Machine Learning is a subset of AI. When we say be intelligent to a machine, we need to construct a certain framework through which the abstraction can be put to work. Machine Learning is a way in which we train the Machines to Learn about the General and Specific needs of humans.

• Through (Knowledge Management System & Expert Systems) we impart Facts to the system, Domain knowledge, and also the rules to react to situations.
• Through Statistical Models and ML Algorithms, we aid in training on data, analyzing patterns in Data, and Tuning attributes to make future predictions and aid in the decision-making process.
• Through Big Data, we handle data characterized by high Volume, Velocity, and Variety, for storage, processing, and analytics) to build AI Systems. (There are so many Data pipelines, Data Analytics Tools, and AI applications, but at a foundational level the above 3 are what constitute systems building at a bigger scale)

How can we broadly classify Machine Learning?

Different technological and business needs mandate for classifying Machine Learning algorithms. Broadly classified as Supervised and Unsupervised learning. Other classifications are refined from this basic classification to meet specific business/technological needs.

In Supervised learning, the learning algorithm is trained on a labeled dataset, where each input is paired with its corresponding output. The algorithm learns to map input data to the correct output by generalizing patterns from the labeled dataset. Some of the common Algorithms that fall under this are (Logistic Regression, Naive Bayes, SVM, Neural Networks...etc.).

Alternatively in Unsupervised learning, the learning algorithm is given an unlabeled dataset and must find patterns or relationships within the data without explicit guidance. Some of the common algorithms that fall under this are (K-means, Hierarchical Clustering, DBSCAN..etc).

Algorithms like Neural Networks are tricky as they could be used as Supervised/Unsupervised formats, so it depends on the specific tasks we try to attain. Many of the AI applications today use an ensemble of different ML Algorithms to validate various statistical parameters and arrive at desired business outcomes.

How has Data Science as a field evolved?

Finally, the field of Data science is quite important for the success of both ML & AI with the Nature and scale of Data we are dealing with in this digital era.

The field of Data Science (DS) derives from various disciplines, but more importantly traced back to Statistics, Computer Science and domain expertise (from various industry verticals). Along with the above basic foundations, Information theory, Data Mining, and Business Intelligence, Bigdata & ML have given more prominence to Data Science to evolve rapidly.

• Information theory, developed by Claude Shannon in the mid-20th century, has influenced data science. Information theory provides a framework for quantifying information and uncertainty.
• Data mining involves discovering patterns and knowledge from large datasets. It played a role in the evolution of data science, especially in uncovering valuable information from data.
• Business intelligence practices, which involve collecting, analyzing, and presenting business data, contribute to the foundations of data science.
• The Explosion of digital data, coupled with advances in storage and processing power, led to the emergence of Big Data. Technologies like Apache Hadoop and distributed computing became essential for handling massive datasets.
• The widespread adoption of Machine Learning techniques, fueled by improved algorithms and increased computing power, became a defining aspect of data science.

The domains of AI, ML, and DS are advancing swiftly, arguably outpacing the traditional trajectory of technological progress defined by Moore's Law. It's crucial for us all to continually learn and grow within these dynamic fields.