ML and DL help us to create an AI system that can even create systems with more complex tasks that a human can do
Artificial Intelligence (AI) is a science that deals with how we can build a system using computers, and the data that we have in the humongous form, that can do the manual tasks that humans do in to an automated one. Robotics Process Automation (RPA) deals with the conversion of manual process into automation. AI is still bigger than just automating the computer processes. It also has ML and DL at a very greater scale. ML and DL help us to create an AI system that can even create systems with more complex tasks that a human can do. The way human analyzes a situation and behaves or works here at first stage we want to see that the way a reasonable person with average understanding can do, such tasks can be done using ML and DL in an appropriate way.
In this article we will see some of the examples that can show us how AI can be powerful and its underneath technology, using ML and DL and the different formulae applied and then the power that AI gets to show how it can help humans in automating many things. Certainly, AI must be explored more and with the extent that AI has, it can help humans while they do some tasks manually and the human intervene deciding how to take this ahead.
I was going through the videos and articles that discusses many examples that define the real-life AI and had done good things that make our life easier. When we see in our daily lives there are so many tasks that are repetitive and are done manually by people, even as a layman we would think if we can get some automated solution for these manual tasks to help us in a big way. The work that we do manually in our computers like data entry operations, or any support operations and try to make those into an automation process. AI is intelligent enough to help us in solving these sorts of manual repetitive tasks in the start and head towards being the leader in automation industry. This simply transforms reading of mails and other files and replying to those mails with not just auto-reply sort of things, but an added intelligent in the reply using ML technology.
Data Science and AI show us how important the need of Automation is in today’s world. When we do so many of the repeated tasks manually and getting connected from one software to another, and sometimes from one server to another and repeatedly doing those all the tasks manually, drain all our energy, and that is the time we hope of some such automation process if it exists, we will welcome it with our open hearts. This is only the striving thing that wants us to think and create a solution that would help the techies to just click a button and all the tasks are done automatically. For example, take a scenario, if there are some unread emails of the same subject, pending to be read for 50 days old mails, same emails, with same issues containing same data, but they must go through those emails and figure out when the problem arose and when it was resolved. This can be done only by manually visiting those emails and checking one by one, and finally figuring it out when it was finally resolved so that we can send the closing mail to the recipient. Imagine if we create a bot that goes through these emails and searches all the emails with a same subject line and finally gives out the report saying that this email has been coming from last 60 days and the solution is yet to be done, and this report goes to the SME who is responsible for these emails, will surely make the life easier for the SME, with such an automation solution. Isn’t it?
In the above example, NLP (Natural Language Processing) helps us in a great way. With its help we can find the emails that contain same subject and then fetch the total number of emails that are there for that subject. The concepts of NLP like finding a word in a paragraph and by applying stop words we can eliminate words like pronouns, conjunctions, etc. to get a clean paragraph and then even execute other concepts to figure out the specific word from a paragraph or an article to fetch more required information. For example, if we want to search the internet for a celebrity where all and which all magazines and/or newspapers his name is mentioned or for which brands he is advertising, this can be very easily fetched through NLP and ML concepts of unstructured data.
ML opens up more ways to automate the process by applying concepts in AI. The two most important concepts are:
- Supervised learning
- Unsupervised learning
Easier and powerful supervised learning in which we have a last column called result, and other columns called fields that can sort out the values and the data in a particular fashion.
For example, if we have a training data of one of the humongous data sets of the “breeds of dogs” which the machine is made to recognize based on its features and the machine has now learnt much from its training data that is fed. Now when it sees a picture of a dog in the real time data, input check, it goes through all of its training data set and based on that it can recognize which dog it is. In the machine learning this accuracy comes up to 70% to 75% and that is considered as the best accuracy. But same thing when it gets into deep learning (using CNN - Convolution Neural Network) then the accuracy shoots up to 95% because now the machine is doing its analysis like a human brain does.
These and similar examples show how AI interact with automation and help us in our day-to-day activity. It makes these technologies go a long way to make more and more intelligent solutions for us so that not only repetitive tasks but also apart from these we get more better solutions. We expect these technologies to play a vital role in displaying its presence in these manual processes and show more enhanced behavior.
The author is Manager - IT at Atos India