When we combine domain expertise and scientific methods (Maths and Statistics) with technology (Python, R, other languages) we get Data Science. Data scientists collect data, clean the data, explore the data, analyze the data and visualize the data to find patterns. They apply mathematical and statistical models to make predictions and find solutions to problems.
Understand the problem with the help of a domain expert.
For example a restaurant wants to increase it's price list. It has to increase in such a way that the customers will not stop coming to the place because of over pricing yet generate increased profit. Think of all the data that the data scientist can use. The data scientist will have to collect all the data which he thinks will help in addressing this issue. The relevance and irrelevance of data is done scientifically. So at the outset he/she just takes all the data available.
There are 3 types of analysis -
- Descriptive analysis - This is where the data scientist gets complete understanding of the data.
- Predictive analysis - Based on the data the data scientist makes predictions.
- Presecriptive analysis - The data scientist prescribes some changes based on the data.
In this case we are expecting the data scientist to prescribe the % of price increase and may be items which can be increased.
The Data scientist will develop a learning model to make the prediction and prescription. So in this case the data scientists work ends once he prescribes.
Machine learning is when data scientist analyzes the issue, creates a model into which data is fit and as and when there is more and more data available we keep fitting it into the model.
Take this analogy as an example. You want to buy an item. You call and check with the shop on Friday. They say they are open from 9 to 6. That is the data you have. Saturday morning 9 you go there. It is closed. You find out they are open only from 10 am on Saturday. Based on the new data you will change you plan of action and from next time you will plan to come only at 10 if you come on Saturdays. Then when you buy the item, you find out that you can only pay cash. You are not carrying cash. That is again additional data available. So now if you ever go to the shop, you will use all the data you have make your visit successful.
Same way with machine learning with each new data coming in, the algorithm will change automatically and the prediction result will vary depending on the data.
But the machine learning model needs to be given the new data to learn. And the model needs to be used within a system for it to be harnessed. The process of building machine learning into a system makes it artificially intelligent.
Imagine Siri or Alexa or Google assitant - These are all AI enabled chatbots. The first time you buy a phone and ask Siri what your name is, it may not know. Based on text, it will pronounce your name. Then you say the right way to say the name and it learns. It appears to be a simple process. But actually a lot of processes are happening. The machine is converting your speech to text. Understanding it . Answering it based on resources available. And then render it as a voice message again using text to speech.
An AI enabled system may have one or more machine learning and/or deep learning components. The program which passes the data to the AI system for further learning is also an integral part of the system.
Use the existing AI tools to create solution. There are many companies which provide tools to create AI solutions. To implement AI in a company, we can either use these tools or build from the scratch. If the purpose of AI is a generic one, it is econimically a better decision to use the readily available tools. For instance, to build a full fledged chatbot for an organization, it requires the efforts of many engineers and testers who may take a long time. And then there is need for enhancement and maintenance to adapt to evolution of technology. It makes sense to use tools available like IBM chatbot and then customize it to an organization's need. To use AI tools you need to be aware of the tools and how to configure them. This is what the learner learns in Applied AI.