Definition: AI in a nutshell

AI is enabling dramatic and positive breakthroughs in science and medicine, but it is a disruptor, and you hear warnings all the time. To learn why experts are fearful, see AI anxiety.

Pattern Recognition and Robotics
AI is essentially two things: pattern recognition and robotics. Pattern recognition identifies images, natural languages, trends and behaviors. Robotics are mechanical devices with articulating limbs that take the form of two-legged humans and four-legged animals. However, most robots are industrial devices of all shapes and sizes with one or more extremities. Nothing to cuddle up to, but industrial robots are indispensable in manufacturing and logistics. Robots also employ pattern recognition, an essential component. Their machine vision and language skills enable them to recognize objects and converse in a human language. See robot and AI.

What AI Does
In medicine and science, AI is used to explore the interactions in every molecular structure. AI is used to make marketing and financial forecasts, and AI is used to answer questions about anything. A huge step beyond the voice assistants in every phone, AI systems also provide the extraordinary service of generating content that most people think was created by humans. ChatGPT and similar AI chatbots can create essays, articles, images and videos. Ask an AI to come up with a poem about a frog and a pickup truck, and it will create an eloquent one. Unfortunately, it is increasingly difficult to tell the difference between human-created and machine-created content (see deepfake). See generative AI, ChatGPT, GPT and Gemini chatbot.

The Tools - Not Ordinary Programming
Nothing like ordinary application programming, AI uses a "neural network" made up of multiple layers that connect to each other. There are three parts to an AI system. The programming phase sets up and fine tunes the network; the training phase feeds it the data to learn about a subject and the execution phase, known as "inference," processes the results. To analyze the patterns, the training takes in billions and trillions of data samples using pre-identified (labeled) content or unidentified content. See AI data labeling and inference engine.

Known as "language models," the more layers in the neural network, the "deeper" the learning. The more samples of data fed in the training stage, the larger the knowledge base and the more comprehensive the results when asked to do work (analyze; predict; generate). See deep learning.

The data samples come from the Internet, including websites. blogs, articles, dictionaries, encyclopedias and books. Essentially the world's information. Training phases can take a huge amount of datacenter time, power and electricity, and specialized graphics processing units (GPUs) are used (see GPGPU). Microsoft plans to reopen the Three Mile Island nuclear power plant in the center of Pennsylvania to generate electricity for AI processing.

A Simple Example
An easy-to-understand example of how AI pattern recognition is used in medicine is x-ray analysis. If 10,000 chest x-rays showing lung cancer and 10,000 cancer-free x-rays are fed into a neural network, the system learns the differences between them. Such systems can detect diseases as accurately or better than medical professionals, and most importantly, faster. See large language model, deep learning and neural network.

Everything Is a Pattern
No matter what people do in life, over time, they perform repetitious patterns. When these patterns are captured as data, they can be used to train an AI. Although human intelligence is implied, AI results are the regurgitation of historical patterns combined with varying degrees of influence programmed into the neural network algorithms by AI designers.

Concern for the Future
What worries people is the research being done to replace human decision making with AI. For example, should AI be used to expand a company or pull back? Even more significant, do we leave the decision to go to war up to a machine? There is a huge amount of controversy regarding AI and the future (see AI anxiety). See AI, AI hallucination, AGI and AI stages.

AI Will Reign Supreme

MIT professor Max Tegmark's best-selling book postulates an AI that far exceeds human intelligence and literally takes over.

A Note from the Author
Everyone has an opinion about AI whether they really understand it or not. Well, so do I... after all, I've been in the industry more than 60 years and have made more than a hundred thousand edits to the content you are reading.

If AI replaces human analysis, it may cause immense problems, especially if a decision to go to war is followed. However, I believe we are suffering more misery due to media disinformation that user engagement algorithms, whether AI-based or not, are increasingly making worse. Already here and now, outright lies everywhere may prove far more dangerous to civilization than robots taking over the world! See disinformation and user engagement.