What artificial intelligence and machine learning exactly is ?
AI stands for artificial intelligence, where intelligence is defined as the ability to acquire and apply knowledge.
ML stands for machine learning where learning is defined as the acquisition of knowledge or skills through experience, study, or by being taught.
Imagine we want to create artificial ants who can crawl around in two dimensional space. However, there are dangers in this world: if a ant encounters a poisonous area, it will die. If there are no poison in ant’s proximity, the ant will live.
How can we teach ants to avoid poisonous areas, so that these ants can live as long as they wish? Let’s give our ants a simple instruction set that they can follow; they can move freely in two dimensional space one unit at a time. Our first attempt is to allow ants to crawl around by generating random instructions. Then we tweak these ants and let them crawl around the world again. We repeat this until ants successfully avoid the poisonous areas in the world. This is a holistic machine learning way to approach the problem. We make ants to fit in configuration by using some arbitrary rule. This works because in each iteration we prune away a set of non-fitting ants. Eventually, we are pushed towards more fitting ants.
But what if we change the location of poisonous areas, what do you think will happen? Ants would undergo a huge crisis because they couldn’t survive in the world anymore – they couldn’t simply know where the poisonous areas are and therefore would not be able to avoid them. But why this happens, and could we do any better? Could ants somehow know where the areas are and adapt their behavior to make them more successful? This is where artificial intelligence comes into play. We need a way to give ants this information, give them knowledge of the environment. Our ants need a way to sense the world. Until this, they have been living in completely darkness, without any way to perceive the world around them. For example, we can let ants to leave a short trail which other ants can sense. Then we can make ants to follow this trail and if they cannot sense a trail, they just crawl around randomly. Now, if there are multiple ants, most of them will hit the poisonous areas and die. But there are also ants who won’t die and therefore crawl in a non-poisonous areas – they will leave a trail! Other ants can follow this trail blindly and always know that they will live. This works because ants can receive some information of their surroundings. They can’t perceive the poisonous areas (they don’t even know what poison is), but they can avoid them even in completely new environments without any special learning.
These two approaches are quite different.
- The machine learning way tries to find a pattern which ants can follow and succeed. But it doesn’t give ants a change to make local decisions.
- Artificial intelligence way is to let ants to make local decisions to be successful as a whole. In nature, we can find many similarities to this kind of artificial intelligence way to solve problems.
Hope that draws clear distinction between AI and ML.
Comments
Post a Comment