Overview
- Artificial Intelligence (AI) is the broadest concept, encompassing machines or systems designed to simulate or mimic human intelligence.
- Machine Learning (ML) is a subset of AI that focuses on systems that can learn from and make predictions or decisions based on data. Instead of programming explicit rules, ML systems learn patterns from data to make predictions or decisions.
- Deep Learning (DL) is an advanced form of Machine Learning that uses multiple layer of Neural Networks. It’s very powerful when you need work with complex pattern recognition.
- Data Science it’s a disciplines that intersects AI and ML. The Data Scientist use statistical and mathematical methods, ML techniques and domain expertise to extract insights from data. The Data Science might use AI/ML tools, but their goal is to analyze, understand and make decisions with the data, rather than build AI systems.
- Computer Vision and Natural Language Processing are applications domains within AI. They’re specfic problem areas where AI techniques (including ML and Deep Learning) are applied.
Different Approaches between techniques
The key distinctions between Machine Learning and Deep Learning lies in their approaches to create “intelligent” behavior.
- AI as the broadest category encompasses multiple approches like:
- Rule based systems: Explicitly programmed “if-then” rules (e.g. “if this to that).
- Expert systems: Codified human expertise into logical rule.
- Search algorithms: Chess programs evaluates possible moves.
- and Machine Learning: Learning from data (which is just one approach within AI)
TIP
Key point: You can create AI without any learning - just good programming and logic.
-
Machine Learning requires data to learn patterns (it’s data-driven)
- Instead of programming rules, you provide examples (data)
- The algorithm finds patterns in the data
- It uses those patterns to make predictions of new, unseed data
- No data = no learning = no ML
-
Deep Learning need large amount of data and computing power (resource-hungry)
- Large datasets: Often need millions of examples to work well
- Computing power: Training neural networks with many layers requires significant processing (GPUs, cloud computing)
- However they have better performance on complex tasks (like image recognition, language understanding)