Artificial intelligence (AI)
The broader goal of creating computer systems that perform tasks associated with human intelligence — such as understanding language, recognizing images, planning, reasoning, or making decisions.
Basic distinction
AI vs. ML
Two terms that get used almost interchangeably, but describe different things — one is a goal, the other is a technique for reaching it.
Machine learning (ML)
One way to build AI systems. Instead of programming every rule manually, developers train an ML model using data so it can recognize patterns and make predictions.
AI is the overall capability; ML is a technique used to create that capability.
Not every AI system uses machine learning — an AI system could rely entirely on hand-written rules. Most modern AI products use ML heavily, but the two words aren't synonyms.
Example: identifying spam email
The same problem, solved two different ways, shows the distinction clearly.
Rule-based AI approach
- If the subject contains “You won!”
- and the email contains several suspicious links,
- move it to the spam folder.
A developer manually defines those rules.
ML approach
- Provide thousands of emails labeled “spam” or “not spam.”
- The ML system learns patterns from those examples.
- It predicts whether a new email is spam.
The finished spam filter is an AI application. The pattern-learning process behind it is machine learning.
AI tools vs. ML tools in the cloud
Cloud AI tools hand you a finished capability. Cloud ML tools hand you the infrastructure to build your own.
AI tools
Ready-to-use intelligent capabilities — send data, get a result. The provider has usually already trained the underlying model.
- Speech recognition: Convert a recorded customer-service call into text.
- Text-to-speech: Generate a spoken version of an article.
- Language translation: Translate an English support message into Japanese.
- Image recognition: Identify objects, faces, or text in an uploaded image.
- Chatbots: Answer customers' questions using natural language.
- Document analysis: Extract names, dates, totals, and addresses from invoices.
- Generative AI: Create text, summaries, images, or code from instructions.
ML tools
Generally used to build, train, evaluate, deploy, and monitor custom models:
- Preparing and labeling training data
- Selecting an ML algorithm
- Training a model on cloud servers
- Testing how accurate the model is
- Adjusting model settings
- Deploying the model as an API
- Monitoring whether its predictions become less accurate over time
A retailer might use ML tools to train a model that predicts:
- Which products a customer may buy
- How much inventory will be needed next month
- Whether a transaction might be fraudulent
- Whether a customer is likely to cancel a subscription
Side-by-side examples
The same situation, approached as a consumed AI service versus a custom-trained ML model.
| Situation | AI tool | ML tool |
|---|---|---|
| Analyze customer reviews | Use a ready-made service to identify positive or negative sentiment | Train a custom sentiment model using reviews from your industry |
| Detect objects in photos | Use a service that already recognizes cars, people, and buildings | Train a model to recognize a company's specific products |
| Predict fraud | Use a prebuilt fraud-detection service | Train a model using the company's historical transaction data |
| Build a chatbot | Use a conversational AI service | Train or fine-tune a model using company documentation |
| Forecast sales | Use an automated forecasting service | Build and test a custom forecasting model with selected algorithms |
A practical analogy
Think of AI as a finished meal and ML as part of the cooking process.
- An AI service is like ordering a prepared meal. It is ready to use.
- An ML platform is like having a professional kitchen, ingredients, equipment, and tools so you can create your own meal.
The terms sometimes overlap because an ML platform may also include ready-made AI services. Course material often uses “AI/ML tools” as an umbrella phrase covering both: ready-to-use intelligent services, and tools for developing custom machine-learning models.
Related terms
See AI, ML, inference, and fine-tuning in the Glossary.