AI (Artificial Intelligence)

• AI = Artificial + Intelligence.

• Artificial = Man-made or created by someone.

• Intelligence = common sense or generalized algorithm etc. which acts like a human.

• It was started in the 1950s by "Alan Turing".

• The first AI was "Logic-based AI" (1950 - 1960).

• The Second was "Rule-based AI" (1960 - 1970).

• At this time there was a boom in technological advancement, which led to "Symbolic AI".

• It is built with if else conditions.

• Symbolic AI resulted in the emergence of Expert systems.

• Symbolic AI when used in a specific domain is called Expert System.

• These emulate the decision-making of a human expert.

examples are Chess players etc.

• Flaws in Expert Systems

• Focused only on specific closed problems, like lung detection. chess players etc.,

• It won't help in Fuzzy Logic, like predicting a dog in a picture.

So, expert systems got left behind. Later, ML came into the picture.

ML (Machine Learning)

• ML = Machine + Learning

• Machine = Robot or any other computer.

• Learning = the process of training a computer program to make predictions based on input data.

• It is a branch of "AI".

• It uses statistical methods to find patterns in Input data.

• No need to explicitly program.

• "ML" is better than "Symbolic AI".

• Needs lots of data.

• learning = recognition of patterns in data.

• No need to write rules like SymbolicAI just input lots of data, it will generate its own rules by the data.

• ML is inspired from statistics.

  

If Machine Learning is doing great. Why need for Deep Learning?

DL (Deep Learning)

• DL= Deep + Learning

• Deep = Having many steps of calculations.

• Learning = Recognizing patterns in data.

• DL is also Machine Learning.

• But algorithms are different.

• DL is inspired by human biology but DL won't work as the human brain.

Why Deep Learning?

• ML cannot do some things well enough.

• In, ML Input features are provided.

• But DL detects/extracts features on its own from the data.

• In fuzzy logic, we don't know the features, but DL helps you to predict.

• example: Predicting placement is done(Yes or No) from the data(Resume).

• More layers increase the accuracy or efficiency of the model gradually.

• More data = more performance of the model for DL. But ML stabilizes after a certain point.

DL is not used everywhere. It is used I working with large data.

If data = Images or videos or sequences. DL is used.