AI Concepts Made Easy: A Beginner's Manual (Part 1 of 5)
Artificial Intelligence (AI) is a broad field that aims to create intelligent machines capable of mimicking human cognitive abilities. AI systems are programmed with human-defined rules to perform specific tasks. However, AI relies heavily on human intelligence for setting rules and boundaries. Some of the AI examples include -
A robot that can solve a Rubik’s cube with pre-defined rules given to it by human programmers.
software that could read cheques.
fraud detection systems that uses rules to flag abnormal spending patterns.
Machine Learning (ML) takes AI to the next level by enabling computers to learn and improve without explicit programming. ML algorithms learn by analyzing large historical datasets to find patterns and build models that can make predictions. The key difference is that ML algorithms/models self-improve based on experience, without human intervention.
For example, an ML model can be shown thousands of lion and fox photos, and it will learn the patterns itself to distinguish between the two. When presented with a new photo, the model can predict if it's a lion or fox based on its prior learning.
ML models can process exponentially more data and get more accurate over time as they process more data. The ability for ML models to ingest exponentially more data than humans makes them powerful for unlocking insights and patterns that would otherwise be impossible to uncover. Additional examples would be -
Stock market predictions
Predict correct translation from one language to another
Image recognition in smart phones
Music and movie recommender systems like Spotify, Amazon, Netflix that learn your preferences and suggest songs & shows based on complex patterns learned from millions of users.
There are two main types of ML algorithms:
Supervised learning trains models on labeled datasets, where inputs are mapped to known outputs. The model learns the relationships between inputs and outputs. The goal is to produce accurate predictions for new, unseen inputs based on these relationships.
For example, it's like a parent teaching a toddler what to do in specific situations. The parent provides examples of inputs (situations) and the desired outputs (proper responses). Through consistent feedback and labeling of the training examples, the toddler learns how to respond appropriately to new situations.
Additional Real-World applications of Supervised Learning -
Email spam filters are trained on thousands of labelled emails as spam or not spam. The model learns to classify new emails.
Similarly, Netflix collects your ratings and feedback on movies and shows to train their recommendation models. The more data they collect on user preferences, the better their models become at predicting shows that you are likely to enjoy watching.
Unsupervised learning is a model that is self-trained on large datasets, trying to find patterns and structure without any labeling or guidance. These are deep learning models that rely on the density and distribution of data to learn. The open-ended nature of unsupervised learning can lead to novel discoveries and creativity.
Unsupervised learning is like a parent giving a child toys and books and saying "Have fun and figure out what to do with these." The child gets to independently explore and learn by discovering patterns, grouping similar items, and making their own connections.
Additional Real-World applications of Unsupervised Learning -
Segmenting customers into groups based on purchasing patterns.
Grouping songs by musical features like rhythm, tempo, instruments
In summary, AI aims to create intelligent machines or software, while ML is a method for achieving AI using data-driven predictive modeling and improving based on experience. ML takes AI to the next level by enabling machines to learn without explicit programming.
Stay tuned to learn deep learning in my next article.
Happy learning!
Sources: