Beyond the Code: Understanding How Machines Learn and Grow

 By: Festus Ewakaa Kahunla

Introduction

Photo generated by author using Dalle-3: Prompt "A robot predicting a child's future"

Welcome to a journey into the world of Artificial Intelligence (AI), a journey that's as relevant in the streets of Freetown as it is in the tech hubs of Silicon Valley. In Sierra Leone, just like elsewhere, AI is no longer just a buzzword but a reality touching various aspects of our lives. But with its rise come many myths and misconceptions. For instance, you might have heard someone express fear about AI, suggesting it could predict a child’s future – a claim that seems more like science fiction than reality. This blog aims to demystify AI, especially focusing on how it learns, to dispel such fears and bring a realistic understanding of what AI can and cannot do.

Understanding AI's learning process is crucial. It helps us appreciate not only its potential but also its limitations. Contrary to some dramatic predictions, AI is not an all-knowing oracle. Rather, it's a tool crafted by humans, learning and evolving based on the data and instructions we provide. By exploring how AI learns, we can better understand its role in our daily lives – from the apps we use to the services we rely on – and why it's not something to be feared but understood and used wisely.

So, whether you're a student in Bo, a business owner in Kenema, or just someone curious about the digital world, this guide is for you. Let's embark on this enlightening journey together, making AI learning not just informative but also fun and relatable

What is AI Learning?

AI learning is like teaching a computer to think and make decisions on its own. Just as we teach children how to identify fruits by showing them apples, bananas, and oranges, AI is taught to recognize patterns and make predictions. For example, showing a computer many pictures of cats helps it learn to identify what a cat looks like. This process of learning allows AI to perform tasks without being explicitly programmed for each step, making it different from traditional computer programs which can't learn from new data.

Contrasting AI Learning with Traditional Programming

Imagine you're teaching someone to cook using a recipe book. Traditional programming is like following a recipe exactly, step by step, without any room for improvisation. The computer can only do what the recipe (or program) tells it to do.

AI learning, on the other hand, is like teaching someone to cook cassava leaf by taste and intuition. Instead of following a specific recipe, the AI learns the basics of cooking cassava leaf and can create dishes it's never made before like groundnut soup, potato leaf and all. It's flexible and can handle surprises, like if certain ingredients are missing. This flexibility makes AI unique, allowing it to adapt and make decisions in situations it wasn't specifically programmed for.

Machine Learning The Backbone of AI Learning

Machine Learning (ML) is what makes AI seem smart. It's like the brain of AI, giving it the ability to learn from experiences. Imagine AI as a student and ML as the learning method it uses to get smarter. Just as a student learns and improves over time, AI systems use ML to continuously get better at doing tasks.

Types of Machine Learning

Photo credit Reddit: Types of machine learning

Machine Learning can be like different styles of teaching. There are three main types, each with its own unique approach:

1. Supervised Learning: Picture a classroom where a teacher provides students with the correct answers as they learn. In supervised learning, AI is given data (like pictures) and the answers (labels). For example, an AI system is shown thousands of photos, some labeled 'cat' and others 'not cat.' It uses these examples to learn how to identify cats in any picture.
2. Unsupervised Learning: Now imagine a student learning on their own by observing the world around them. This is unsupervised learning. The AI is given data but no answers. It has to figure out patterns and organize the data itself. For instance, it might get a bunch of news articles and learn to sort them into categories like sports, politics, or entertainment, without being told what these categories are.
3. Reinforcement Learning: Think of this as learning by doing and getting feedback. It's like training a pet with treats. The AI tries something (makes a decision) and gets rewards or penalties based on the outcome. If the decision is good, it gets a 'treat' (reward), and if not, it gets a 'no treat' (penalty). Over time, the AI learns to make better decisions to get more treats.

Each of these types of learning helps AI systems to handle different tasks, whether simple or complex. By understanding these basics, we can better appreciate how AI is growing and changing the way things work in fields like healthcare, finance, and even our daily use of technology.

Stages of AI Learning

AI learning can be broken down into several key stages:

1. Data Ingestion: This is where AI starts its learning journey. It's like gathering ingredients for a recipe. The AI system collects large amounts of data - images, texts, numbers, etc.
2. Pattern Recognition: Next, the AI looks for patterns in the data, much like finding a recipe that matches the ingredients. It uses algorithms to detect these patterns, which could be anything from recognizing a face in a photo to understanding speech in a voice command.
3. Model Building: The AI then builds a model, a bit like writing down the recipe. This model represents what the AI has learned about the patterns in the data.
4. Testing and Learning: The AI tests its model, much like tasting a dish to see if it needs more salt. It uses new data to see how well its model works, learning from its successes and mistakes, and adjusting the model accordingly.

Over time, this process helps AI refine its understanding and decision-making abilities. It gets better at predicting outcomes and making choices based on the data it has learned from.

AI Learning in Various Systems

Photo of a local translator app developed by author

1. ChatGPT: It learns from a vast database of text. Through supervised and unsupervised learning, its been taught to understand language patterns and respond to questions. Every interaction helps it improve, making it more conversational and accurate over time.
2. Language Translators: These tools use supervised learning. They are trained on large datasets of bilingual text to understand how words, phrases, and sentences translate from one language to another. As they encounter more text and user feedback, their translations become more accurate and nuanced.
3. AlphaGo: This AI program, designed to play the board game Go, uses reinforcement learning. It learned strategies by playing numerous games against itself and improved by analyzing the outcomes of each move, much like learning from each game played.
4. Tesla Cars: Tesla's self-driving cars use a combination of supervised and unsupervised learning. They constantly collect data from their environment (like road conditions, obstacles, and traffic signs) and use it to make real-time driving decisions. This ongoing process allows them to adapt and improve their driving algorithms continuously.

Real-World Benefits of AI Learning

Photo credit Brooke Steinberg: AI develops cancer treatment in 30 days

1. Healthcare: In healthcare, AI systems analyze medical images like X-rays or MRIs, learning to detect signs of diseases such as cancer more accurately than ever before.
2. Finance: In the financial sector, AI algorithms predict market trends, helping investors make informed decisions. They also detect fraudulent activities by recognizing abnormal patterns in transaction data.
3. Personal Assistants: Virtual assistants like Siri and Alexa continuously learn from user interactions. They improve their ability to understand natural language and user preferences, making them more helpful and personalized over time.

Conclusion

In our journey through the world of AI, we've seen its remarkable capacity to learn and adapt, much like a curious student. From the initial stages of data ingestion to the complexities of decision-making, AI is reshaping our lives, enhancing everything from our smartphones to the advent of self-driving cars.

As AI continues to evolve, it's natural to have concerns about its impact. In these moments of uncertainty, it's important to recall the words of Moses Joseph Lappia: "Fear the people than the tool." This poignant reminder emphasizes that the direction and use of AI are in human hands. As machine learning engineers and guardians of this technology, our role is to steer AI towards safe and ethical applications, ensuring it serves as a force for good.

Let's approach AI with an open mind and a sense of curiosity. Understanding AI is not just about recognizing its technical capabilities but also about appreciating its potential to enhance our world. As we learn and grow with AI, we have the opportunity to shape a future brimming with possibilities. So, let's fear less and explore more, remembering that the future of AI is not just about how machines learn – it's about how we, as a society, learn to use them wisely.

Together, let's embrace this extraordinary journey with AI, guided by insight and responsibility, to forge a path that is as remarkable as the technology itself.

Photo generated by author using Dalle-3: Prompt "AI leading us to a better future"