Application of Machine learning in Artificial Intelligence
Artificial Intelligence and Machine Learning are the two most trending topics nowadays among computer enthusiasts. These two terms are correlated to each other a lot, even some people use these two interchangeably. Artificial Intelligence is the concept of making machines, computer systems smart which simulates the human brain behavior by injecting its thinking ability, decision-making capability into a computer system, while machine learning is making computer systems learn by themselves using past set of data without the need of explicit programming. You may think AI and ML are the same with different explanations. But it’s wrong! AI represents our brain. Our brain does not operate solely based on past data. So is AI. In a nutshell, Machine learning is a part or subset of Artificial Intelligence. AI is much broader than Machine learning.
Artificial Intelligence systems use algorithms that can work with their own intelligence. It also involves some Machine Learning algorithms such as Reinforcement learning algorithms, deep learning neural networks, etc.
Reinforcement Learning is one of three main paradigms of Machine learning also applied in Artificial Intelligence. It can be interpreted as a set of decisions made by an AI(let’s name it as an agent) to achieve a goal in an uncertain, complex environment. It’s a gamelike situation where the agent(or the AI) has to make decisions on the environment and based on that, the agent may get rewards or penalties. The goal is to maximize the rewards. Game rules and reward policy are set by the designer and he does not provide any clues or hints to the agent. The agent has to play on its own. In the beginning, it uses totally random trial and error and makes a dataset based on the results. Then makes decisions based on that dataset and eventually finishing with sophisticated tactics and skills maximizing reward. This is where Machine learning comes into play. To achieve the task, Machine learning helped by analyzing past actions and getting decisions based on that. This is exactly like how we played video games when we were little without knowing any instructions. We do some trial and error and get to know which actions earn us points and which costs us to lose points. Based on that we try to repeat and identify patterns of success tries to maximize points. (P.S:- Can you remember how we played minesweeper when we were little?) Here we were like the AI and the environment is the video game. In contrast to us, Artificial Intelligence can gather experience from millions of parallel gameplays and reach to goal much faster when Reinforced learning is implemented on a sufficiently sophisticated computer.
This Reinforced learning is applied to AI in autonomous driving systems including trajectory optimization, motion planning, dynamic pathing, and scenario-based learning policies for highways. AWS Deepracer is an example of such an AI model designed to test out reinforced learning in a physical track. It uses cameras to identify the environment and a reinforced learning model to control the throttle and direction.
Other than Reinforced learning, Machine learning has two other types, namely supervised learning and unsupervised learning. Supervised learning is done with help of labeled data.it means some data is attached with the expected result. So it can be compared with new data and obtain the expected result with high accuracy.in unsupervised learning, there is no any labeled data. Therefore the Model has to dive in and work on its own to discover information. So unsupervised learning methods is more complex than supervised learning methods. In google search engine it uses these methods to predict our search item. Google Assistant is the AI that google developed with help of supervised and unsupervised machine learning techniques.
These Machine learning and artificial intelligence algorithms demand high computational power than normal consumer computers can provide today. But it is expected that in near future there may be computer systems intelligent enough that a normal family will be able to raise an AI robot as a child. (Did you remember AI the movie?)
0 Comments