In the last lesson, we discussed the definition of AI
But what about many other terms floating around - Machine Learning, Deep Learning, Generative AI, etc? So let us try to understand these and also look at some examples.
Artificial Intelligence
A broad field of computer science that is focused on creating thinking systems that can perform tasks that typically require human intelligence. The range of such tasks could be reasoning, problem-solving, perception, learning, language understanding, and decision-making. So in that way AI is a catch all phrase of everything - Rule Based Program (If Else conventional programming), Expert Systems, Machine Learning, Deep learning, Generative AI, Natural Language Processing (Giving Machines an ability to understand Language), Computer Vision (Giving Machines an ability to see), Robotics (Giving Machines an ability to act).
Broadly Artificial Intelligence can be achieved in two main ways.
Machine Learning (ML)
Machine Learning (ML) can be considered a class of learning techniques used to achieve Artificial Intelligence (AI) goal. ML focuses on creating a thinking machine by feeding machines data. Instead of relying on explicit instructions, as in Rule Based Program, ML Models learn patterns and relationships from labeled or unlabeled data, enabling them to make predictions or decisions based on new data. Examples of machine learning techniques include decision trees, support vector machines, and neural networks.
Deep Learning (DL)
A subfield of machine learning that focuses on training artificial neural networks with multiple layers, hence called Deep neural networks to learn complex representations of data. Deep learning has gained widespread attention and popularity due to its ability to automatically learn hierarchical features from raw data, such as images, text, and audio. Deep learning has achieved remarkable success in tasks such as image recognition, natural language processing, and speech recognition, often surpassing human performance in certain domains.
Note In Machine Learning, Feature Extraction is done by humans while in Deep Learning Feature Extraction is done by Machines themselves? The feature what? Feature extraction. Let me explain with a business application.
Suppose I want to predict at what Price a house in market will sell at.
What I do I need to be able to answer this?
#1 - Configuration or attribute of a house in the above example, in Machine learning is called ‘Feature’ by Data Scientists. Identifying which features are important is called ‘Feature Extraction’.
But if Machines can do Feature selection or Feature Engineering then why let Humans do this?
We will answer that later. But before we need to understand a few more differences between ML and DL models.
| Aspect |
Machine Learning (ML)
|
Deep Learning (DL) |
|---|---|---|
| Feature Extraction |
Features are manually extracted by humans based on domain knowledge.
|
Features are learned from raw data by the model.
|
|
Model Complexity
|
Generally uses simpler models (e.g., decision trees, logistic regression). | Uses complex models with multiple layers (e.g., deep neural networks). |
| Data Requirements | Can learn from the smaller data set | Requires a much larger data set |
| Interpretability | Models are often more interpretable and transparent. | Models may be less interpretable due to their 'Black Box' Nature. |
| Scalability | May not scale well to large datasets or complex tasks. | Scales well to large datasets and complex tasks with appropriate computational resources. |
| Performance | Performs well on structured/tabular data with well-engineered features. | Excels at tasks involving unstructured data (e.g., images, text, audio). |
| Training Time | Faster to train compared to deep learning models. | Longer Training Time due to complexity models and larger datasets. |
| Use Cases | Used in many traditional applications such as finance, healthcare, and business analytics. | Well-suited for complex tasks such as image recognition, natural language processing, and speech recognition. |
So coming back to the question ‘if Machines can do Feature selection or Feature Engineering then why let Humans do this?’
It is because DL requires a really large amount of data as compared to ML, hence much more time intense. The DL technique is also ‘Black Box’ approach i.e. we do not know what features Machine is selecting and why?
Generative AI (Gen AI)
Generative AI is a subset of Deep Learning (DL). Generative AI refers to AI systems that can generate new data that are similar to, but not the same as, the training data. For example - Letting Machine write Texts, generate Images, Create Music etc.
This is why most textbooks use this Venn diagram to explain the relationship between AI, ML, DL, and Generative AI. Now this all makes sense.
.webp)