Artificial Intelligence ! What is it actually ?
Artificial Intelligence or AI as we call it has become a new buzz word in today’s world. It has become a phenomenon which attracts many and leaves many mesmerized with its astonishing accomplishments. On one hand people are awestruck with the major feats such as self-driving cars, alpha-go (which has handsomely beaten the world GO champion 4–1 ), autonomous machines, landing of Falcon 9 , and, on the other some are equally scared by the rise of machines against humans. The concern of AI ruling humans is not only amongst non-scientific community, but, is equally shared by some of the stalwarts like Elon Musk and Prof. Stephen Hawkins.
With an increase in acceptance of machine learning, data science and AI in the society, this post is an attempt to understand what AI actually is, how it’s being used in various fields in the society and what are some common myths attached to it.
What is AI ?
In simple terms AI or Artificial Intelligence means providing a machine the ability to think and take decision. The goal of AI is to provide a machine with an ability to come up with best possible decision, after considering all possible scenarios and their outcomes in a given environment.
On a technical level, the field of artificial intelligence can be broken down into primary three fields
1. Machine Learning
2. Reinforcement Learning
3. Rule based Learning
Machine Learning: Machine Learning is the art of extracting knowledge out of the current data. Basic principle of machine learning is to train an algorithm or a combination of algorithms on a sample data to learn it’s characteristics and then use this trained algorithm, also called as a model to analyze unseen data with similar characteristics/parameters.
Reinforcement Learning: Reinforcement learning is inspired by behaviorist psychology, concerned with how agents ought to take actions in an environment, so as to maximize cumulative reward. The technique is studied in many other fields such as game theory, control theory, operations research, multi-agent systems and others. The core of this technique is to create a model of an agent and then study it’s interactions with the environment. For a simple example consider a situation of a fight between two wrestlers. Here, the wrestlers are the Agents. There model will comprise of a quantifiable measure of their strength,technique,age,height,weight etc.Environment is the arena where they will be fighting. It will be affected with parameters such as surface, weather, size, audience capacity etc.
Now once these models are prepared interactions of these agents with the environment can be performed to predict the winner or at a more detailed level how individual wrestler will be affected with changes in his environment. This is a highly simplistic example to explain the basic building block of reinforcement learning. Actual learning involves creation gf much more complex examples on these lines.
Rule based learning: Rule based learning was primarily used in the early stages of artificial intelligence when machines were programmed to make decisions on the basis of a fixed set of rules. With the advent of more sophisticated techniques and large availability this approach has mainly phased out and become obsolete.
FAQ’s:
1. Is it possible to create an AI which can do all my work without me doing anything ?
Please do not confuse Artificial Intelligence with magic. In the current stage AI is much behind reaching the normal potential of human intelligence leave aside guessing what a human is thinking. Also, the day AI is able to achieve this feat will mostly mark the end of human civilization.
2. Can AI rule humans ?
Yes and No. Theoretically it is possible perhaps. But, the important factor to note is that how powerful an AI becomes is in the hand of humans creating it. Will humans themselves create their destruction is a thing to be seen in distant future. Currently AI is not that powerful to achieve this feat.
3. Is Deep Learning and Neural Networks AI ?
No. AI is a large field and deep learning and neural networks are just one of the approaches being used. These approaches have recently become popular due to the greater accuracy they have achieved with the large amount of data available. However, they are not the only techniques available and also not to be used in every environment.
4. Is AI some kind of complex programming ?
No. In fact the more you learn about AI the more you experience that it is disconnected with the common notion of programming as understood in society. At a high level AI is a set of mathematical models and concepts which enable you to take a decision and derive insights. In fact many of the great contributors in the field of AI have been mathematicians and not computer scientists. It won’t be an exaggeration to say that to excel in the field of AI one needs to have a solid mathematical background.
5. From where can I quickly learn AI ?
There is no quick way of learning AI. A good place to start is to start learning with courses in data structure, algorithms and probability. This can be topped on with graduate level courses in Machine Learning, Reinforcement Learning, Probabilistic Modeling and other courses.
6. Do i need to learn java/python for learning AI ?
AI in itself is language independent. It’s a set of concepts and techniques which can be applied in any coding language. Having said that python and java are two of the most widely used programming languages. Python is specially useful and tailored for machine learning approaches and is very easy to pick up as well.
7. Are AI and Robotics same ?
This is one of the biggest misconception i have come across. No they are not. While AI largely plays the role of a brain in Robots a number of other fields are required for creation of a robot, such as Haptics, Control Systems, Computer Vision. Basically a successful robot is a result of a beautiful co-ordination of mechanical, electronics and computer engineering.
8. Is use of AI is limited to technology companies ?
This is another misconception prevalent. In fact one of the most wide consumers of AI algorithms have been finance industry. A lot of trades nowadays on stock markets are done by machines (AI algorithms) instead of humans. With real time marketing coming in to play, AI is also making great inroads in the marketing industry. Apart from these, medicine, law, public policy making, planning and optimization,sports and music also heavily use AI and algorithms. In fact there are very few fields nowadays which are not touched by AI
9. How is use of AI different in Technology and Business Industry ?
AI has distinctive uses in the field of business and technology. In the technology the major idea is to make a model which understand the nuances of the environment and can take accurate decisions when being presented with similar environment.
In business, however, it is a completely different ball game. In business it is generally required to find out that additional factor which has been responsible for that growth or results. This is a one step further complex problem as it involves not only learning about the environment but also identifying the crucial factors in that environment which are effecting the result.
Thus while AI has been making progress in leaps and bounds in the field of technology in regular business practices it is still highly constrained and requires human intervention at multiple stages to make it actually work.
