Understanding AI

Artificial Intelligence is all around us—helping our phones recognize our voices, suggesting movies we might like, and even assisting doctors in spotting early signs of disease. But for many people, AI still feels mysterious or intimidating. This post breaks AI down into simple, everyday terms so anyone can understand what it is, how it works, and why it matters.

First, let's answer the question, what exactly is AI? In the simplest terms possible, it's when computers perform tasks that normally require human intelligence. A few good examples are recognizing photos, translating languages, and making recommendations. Some examples that you are probably already using include:

• phone assistants such as Siri, Google Assistant
• Spam filters
• Netflix/Amazon recommendations
• GPS navigation and traffic prediction
• Photo apps targeting faces

From these examples you can see that AI is no longer futuristic, it's already familiar to practically everyone. Now, let's dig a little deeper. What are the different types of AI? 

• Narrow AI: does one specific task
• Generative AI: creates text, images, audio, etc.
• Machine learning: a method where computers learn patterns from data
• Neural networks: Loosely inspired by how the human brain works

How does AI "learn" something? There are a few ways. One, by feeding large amounts of data (images, text, recordings...), finding patterns in that data, and making predictions or generating responses based on those patterns and the content of the data. Keep in mind, this is meant to be a simplified explanation of AI. The capabilities of AI are advancing rapidly. One of those capabilities in particular that is quite advanced today, is facial recognition. Facial recognition (FR) has existed since the 1960's. In the 70's to 80's it became automated. In the 90's it went mainstream and in the 2000's commercial use began. Digital cameras and faster computers made FR practical. In the 2010's, AI breakthroughs made it accurate. Deep learning and neural networks revolutionized facial recognition. Systems became good enough for smartphones, social media photo-tagging, and real-time detection. Recognizing and identifying wanted criminals passing through airports isn't just fantasy futuristic fiction for movies any more. It's real. 

Neural networks are a type  of computer system designed to learn patterns, loosely inspired by how the human brain works. They are computer models made up of layers of small units called 'neurons' that work together to recognize patterns in data. They don't 'think' or 'understand' like humans, they simply find patterns and make predictions based on examples they've seen. A simple analogy could be; imagine an assembly line. Layer 1 looks at the raw data, like pixels in an image. Layer 2 identifies simple patterns, like edges and shapes. Layer 3 combines those shapes into meaningful parts, such as eyes or wheels. Layer 4 makes a decision, this is a person, or a car...

Why are they called "neural"? The human brain has neurons connected in networks. The 'neurons in computer 'neural networks' also have neurons, but they are simply math functions, not biological cells.  Think of computerized neural networks as an analogy of the human brain, not a replica. Neural networks 'learn' through a process called training. They look at lots of examples (pictures, text, audio), they make guesses, they get corrected when wrong, they adjust to do better the next time. After countless repetitions they can make accurate predictions. 

Around 2012, bigger datasets and powerful graphics processors made neural networks dramatically better than previous technology. If you haven't noticed, there is a construction booming on for "data centers" worldwide. Unlike distribution centers for goods and services, data centers need not be located near densely populated areas. They are often located in rural, obscure areas. The datasets that these neural networks rely upon have become massive. Never before in the history of mankind has so much information and computing capability been literally, at our fingertips. And it just keeps getting better, every day...