Summary
Highlights
Adon Primo welcomes participants to the first session of the Data and AI program. He provides a walkthrough of the Learning Management System (LMS), explaining how to find learning outcomes, announcements, class recordings, and assignment details. He emphasizes active participation and encourages questions.
The instructor details the assignment report structure, which includes an introduction, task completed, link to code (using Collab and Kaggle), screenshots, and a conclusion. He stresses the importance of sharing Collab links with appropriate access and explains the grading system (out of three marks) and the policy for late submissions, which receive a maximum of two marks. He encourages early submission to allow for redoing assignments if needed.
The course is structured weekly, with new content unlocking each week. Class recordings and slides will be available shortly after each session. Evening sessions cover the same content as morning sessions, offering flexibility. Support sessions are held on Saturdays from 6 PM to 8 PM to assist with assignments, and instructors are available via WhatsApp, email, or LMS direct message.
The session dives into the core definitions of Data Science, Artificial Intelligence (AI), and Machine Learning (ML). Data science is defined as turning raw data into useful decisions, encompassing data collection, cleaning, and analysis. AI aims to create systems that simulate human intelligence, and ML is a subset of AI that allows machines to learn from data without explicit programming. The analogy of cooking is used to explain data science processes, and riding a bicycle for machine learning.
Machine learning is highlighted for its use when rules are complex, data is large, and patterns are not obvious (e.g., fraud detection, recommendation systems). Deep learning, a specialized subset of ML, uses artificial neural networks to learn complex patterns, mirroring how human brains recognize faces. The hierarchy is AI at the top, followed by ML and then deep learning. The relationship between the three is summarized as Data Science producing insights, ML producing predictions, and AI producing actions.
A detailed timeline of AI development from the 1950s to 2025 is presented, including key milestones like the Turing Test (1950), the coining of 'Artificial Intelligence' (1955), and the development of various AI systems like Deep Blue (1997), Siri (2011), and ChatGPT (2022). The concept of 'AI winters' is explained as periods of declining funding and interest due to unmet expectations, largely caused by overpromising, limited computing power, lack of data, and rule-based system limitations.
A discussion on why Africa lags in AI development covers critical issues such as limited funding and investment, shortage of skilled talent, restricted access to large quality datasets (as most data is owned by non-African companies), infrastructural challenges (internet, cloud services, reliable electricity), and low market awareness and adoption. The discussion emphasizes that AI requires significant resources, and local companies struggle to compete and retain talent.
Various job roles in data science and AI are outlined, including Data Scientist, Data Analyst, Data Engineer, Data Architect, Data Manager, Database Administrator, Statistician, Business Analyst, Machine Learning Engineer, and AI/ML Researcher. Their primary responsibilities, required skills, and the tools they typically use are briefly described.
Two major data science methodologies are introduced: CRISP-DM (Cross-Industry Standard Process for Data Mining) and OSEMN (Obtain, Scrub, Explore, Model, Interpret). CRISP-DM is a six-phase iterative framework (Business Understanding, Data Understanding, Data Preparation, Modeling, Evaluation, Deployment) that ensures a structured approach to solving data problems. OSEMN is a simpler, five-step practical framework (Obtain, Scrub, Explore, Model, Interpret) that is easy to remember and flexible across domains.
The application of AI and ML is illustrated across various sectors: healthcare (diagnosis, drug discovery), finance (trading, fraud detection), retail (recommendation systems, inventory management), transport (autonomous vehicles, traffic prediction), cybersecurity (threat detection), education (personalized learning, AI tutors), manufacturing (predictive maintenance, robotic automation), and government/public policy (crime prediction, public health monitoring).
A practical demonstration of web data scraping is provided, outlining the week's assignment. The task involves extracting structured data from a live website using Python on Google Collab. The essential libraries are Requests (for HTTP requests), Beautiful Soup (for HTML parsing), and Pandas (for data storage and manipulation). The instructor shows how to set up the Collab notebook, import libraries, fetch a webpage, extract column headers, and create a data frame. He advises documenting the work, taking full screenshots with timestamps, and ensuring the Collab notebook is shareable with 'anyone with the link' access before submitting the report as a PDF.