AI vs ML vs DL
For todays topic, we start with a simple overview of the topics we see all over, we hear all the time but can get confused quite easily as to how they matter, where they differ or even fit together.
Artificial Intelligence (AI)
First we start with the larger umbrella for all these terms that is Artificial Intelligence, this refers to systems that perform tasks that are normally fit for human reasoning, decision making and creativity, usually these tasks are not pre-defined in nature, as a person, you may want to take a walk in an entirely new area to you but you will effectively your planning, learning and communication skills to traverse this unknown path, avoid danger, avoid harm while accomplishing the initial goal to take a walk and still acting rationally, you do not walk on top of people or into peoples homes just because it is a new environment or you are accomplishing the goal, either works though.
So Artificial intelligence is the concept of a machine that mimics human intelligence, quite a broad concept when you think about how loosely we define intelligence overall.
What then is Machine Learning (ML)?
Machine Learning or ML is a branch of AI that focuses on creating these systems that not only perform tasks but can learn and improve from different experiences. Remember, human intelligence usually requires experiences, learning and improvement over time. Machine Learning seeks to eliminate defining explicit rules to achieve specific outcomes and behaviors.Think about fire, you get burnt once and you avoid it, but it is not an explicit rule, you know under what circumstances you may avoid it, such as a burning building vs lighting a candle, in most cases you light the candle yourself, so there is a fuzzy like match to avoiding fire but not always, confusing but makes sense.
In that aspect, Machine Learning focuses on systems that can learn and improve from experiences automatically without explicitly being programmed, this may be an iterative process done over time, using lots of data - experiences for systems - these systems then use the patterns and features in data to improve and learn.
The final outcome is more intelligent machines that can perform tasks that are inherently harder to explicitly program. Now, using basic human logic, if it was cloudy and it rained yesterday, if I step out today and see clouds, I expect rain, meaning I am learning from yesterdays experience, on an equal note these machines may process data that is more detailed to perform tasks that may be harder to explicitly program, examples include:
- Classification
- Forecasting
- Prediction
- Clustering
And Finally we have Deep Learning (DL)
Deep learning is a bit different; but basically Deep learning is a subset of Machine Learning that makes uses of intricate neural networks based on the human brain, these allow computers to discover patterns and make decisions from large amounts of data.
While in the usual sense Machine Learning makes use of structured well organized or engineered data, deep learning may make use of unstructured data, think of a self driving car, looking at roads, people, objects, other cars, lights and so much more, this is data that you would normally have a hard time defining and putting into a traditional database to define rules that interpret it.
So essentially, deep learning is a subset of machine learning where algorithms learn by using artificial neural networks. One layer may pass information to the next layer and there may be a lot of layers involved to present a solution. Deep Learning is a large leap in AI by closely mimicking the human brain.
All in all, now you may be able to distinguish between these interconnected areas of AI, we will cover more machine learning related concepts and do deeper dives in subsequent articles, thank you for reading!