AI & ML: Get your terms straight

"AI is whatever hasn't been done yet." —Douglas Hofstadter

The terms Artificial Intelligence (AI) and Machine Learning (ML) are often used interchangeably to denote any sort of magic a computer performs that the observing human considers smart or doesn't entirely understand. All startups claim to have innovative AI in their product, but whether that means you should invest in them is a different matter.

Here's the difference of AI and ML, in a nutshell.

Artificial Intelligence is a catch-all term meaning all kinds of tasks a computer performs that are considered to require "intelligence," even if they are entirely based on clearly defined rules. A calculator is arguably an AI because it performs math that a human being would otherwise have had to figure out alone. AI is an old term, originating in the 1950s and was used for all kinds of mathematical and logical tasks that computers were starting to perform, such as playing chess, well before any contemporary uses that would be considered substitutes for human intelligence. It often happens that when a task performed by computers is well enough understood, it's no longer considered AI but just calculation — this is called the AI effect.

Machine Learning, on the other hand, is a subcategory of AI, in which the computer system doesn't require explicit instructions in order to perform a task but can make inferences, detect patterns and even improve its performance autonomously. Applications for ML include image classification and generation, text analytics, language understanding, language translation and language generation, and detecting anomalies in areas ranging from production lines to human behaviour and stock market performance. ML uses algorithms such as neural networks to learn from large amounts of training data and make predictions that can't necessarily be fully explained in a human-understandable way.

To summarise, all ML is AI, but not all AI is ML. Many people use the terms interchangeably and usually when people talk about "innovative AI solutions," you don't know whether they are trying to make their simple rule-based system seem fancier than it is, or whether their solution actually contains ML. However, it's worth keeping in mind that ML is just one way for computers to make decisions, and rule-based AI systems can be valuable on their own and in conjunction with ML. And it's especially important to note that ML is not a magic bullet that can solve all your problems; it's a specific way of solving specific kinds of problems.

More articles in my series on AI fundamentals:

• Part 1: AI & ML: Get your terms straight
• Part 2: Natural Language Processing: Do chatbots have AI?
• Part 3: Supervised and unsupervised learning (coming soon)
• ?Part 4: From prediction to action (coming soon)