AI Terminologies, Benefits and Limitations

Hello students, welcome to Kwickprep. People say AI can do anything. Is that true? Today we learn five key AI words, the real benefits of AI, where it fails, and why AI is not magic.

Let us begin with five words you will hear in every AI chapter. Data means facts and examples, like photos, marks or messages. An algorithm is a step by step method, like a recipe, that tells the computer how to learn. Training is the process where the algorithm studies the data and learns. The model is what we get after training, the learned pattern stored inside the computer. A prediction is the answer the model gives for new data it has never seen.

Here is how these five words connect, drawn as a flowchart. First, we collect data with correct answers, called labelled data. Next, we choose an algorithm, the method of learning. Then we train, so the algorithm studies the data. Training gives us a model, the learned pattern. Finally, we give the model new input, and it makes a prediction.

Let us see the terms in one example, a machine that sorts mangoes. The data is one thousand photos of mangoes. Each photo has a label, ripe or unripe, written by a person. During training, the algorithm learns which colours and spots mean ripe. The result is the model, a mango checker. When a new mango arrives, the model predicts, ripe.

Now pause and predict. Our model was trained only on photos of green and yellow mangoes. Then a red skinned mango variety arrives at the factory. What will it predict? It may guess wrongly, because it never saw a red mango during training.

Now the benefits of AI. AI is fast, so it can check thousands of medical scans in the time a doctor checks a few. It works day and night, so a bank chatbot can answer questions at midnight. It can be very accurate on narrow tasks, like spotting a suspicious bank payment. It handles huge data, like the satellite readings used in weather forecasts. And it helps people, for example by reading text aloud for someone who cannot see.

These benefits reach ordinary lives in India. Farmers can get weather alerts and crop advice on their phones. Students can use learning apps that give easier or harder questions at their own pace. Doctors can use AI tools to screen scans, while the final decision stays with the doctor.

But AI also has limits. It needs a lot of good data, and with too little data it performs badly. If the data is unfair, the model becomes unfair too, and this is called bias. AI has no common sense, so it can miss something any child would notice. Many models cannot explain why they gave an answer, so we call them a black box. And training big models needs costly computers and a lot of electricity.

Here are kinds of failures that have really happened. A company's hiring tool preferred men, because it learned from past hiring records that were mostly men. Chatbots sometimes state wrong facts very confidently, which is called hallucination. Face recognition can work poorly in low light or on faces missing from its training data. Self-driving cars have made mistakes in unusual road scenes they were never trained on.

So why do we say AI is not magic? AI only knows patterns from the data it was given, so it cannot know what it never saw. It works using maths and probability, not feelings or understanding. And humans collect the data, design the algorithm, train the model and check its answers.

Every AI prediction is only a best guess. A weather app may say a seventy percent chance of rain. On that day it can still stay dry, and the app was not broken. That is why important answers, like a medical result, must always be checked by a person.

Let us recap. The five key terms are data, algorithm, training, model and prediction. AI brings speed, round the clock work, accuracy and the power to handle big data. Its limits are its need for data, bias, and no common sense. And AI is maths on data, not magic, so humans must always check it.