A working definition, narrow versus general AI, and where you already meet it.
Artificial intelligence is the work of building computer systems that do things we would normally say take intelligence: recognising speech, translating between languages, spotting patterns, making predictions, and producing text or images.
There is no single official definition, and that is fine. A useful one for everyday work is this: an AI system takes in information, finds patterns in it, and uses them to make a prediction, a decision or a piece of content. Everything else is detail about how it finds those patterns.
Every AI system in use today is narrow. A model that writes fluent emails cannot drive a car; a model that reads X-rays cannot plan a holiday. Each is very good inside the task it was built for and has no understanding outside it.
General AI, a system that can learn and reason across any task the way a person can, does not exist yet. When you read about it, you are reading about a research goal, not a product.
Early AI was built from rules. Experts wrote instructions by hand: if the temperature is above 38 and the patient has a cough, suggest a flu test. These expert systems were predictable and easy to explain, but they broke the moment the world did something the rules had not imagined.
Machine learning turned this around. Instead of writing the rules, we show the system many examples and let it work out the patterns itself. Deep learning is machine learning done with neural networks that have many layers, which is what made modern speech recognition, image recognition and chat assistants possible.
An AI system is not conscious, it does not "know" things the way you do, and it is not neutral. It reflects the data it learned from, including that data's gaps and biases.
Holding on to that one idea will make you a better user of every AI tool you meet.
Comments
Loading comments…