There’s a lot of buzz around Artificial Intelligence at the moment and the term AI seems to be thrown around a lot but what is it exactly? To clear things up, I just want to create an introductory article on this topic.
First of all, let’s look at the definition to avoid confusion. We have to go back to the earliest and hence purest definition of AI, from the time when it was first coined. The official idea and definition of AI was first coined by Jay McCartney in 1955 at the Dartmouth conference. Of course those plenty of research work done on AI by others such as Alan Turing before this but what they were working on was an undefined field before 1955.
Here’s what McCarthy proposed,
Every aspect of learning or any other feature of intelligence can in principle be so precisely described that a machine can be made to simulate it. An attempt will be made to find out how to make machines use language, form abstractions and concepts, solve kinds of problems now reserved for humans and improve themselves.
In essence, AI is a machine with the ability to solve problems that are usually done by us humans with our natural intelligence. A computer would demonstrate a form of intelligence when it learns how to improve itself at solving these problems. To elaborate further, the 1955 proposal defines seven areas of AI. Today they’re surely more but here are the original seven ones:
- Simulating higher functions of the Human Brain.
- Programming a computer to use general language.
- Arranging hypothetical neurons in a manner enabling them to form concepts for.
- A way to determine and measure problem complexity
- Abstraction defined as the quality of dealing with ideas rather than events.
- Randomness and creativity.
After 60 years, I think realistically we’ve completed the language measure problem complexity and self-improvement to at least some degree. However, randomness and creativity is just starting to be explored this year. We’ve seen a couple of web episode scripts short films and even a feature-length film co-written or completely written by AI though they are far from being perfect.
Okay so, in the definition you heard the word Intelligence. What is the Intelligence? Well according to Jack Copeland, who has written several books on AI, some of the most important factors of intelligence are:
Generalization learning: That is learning that enables the learner to be able to perform better in situations not previously encountered.
Reasoning: To reason is to draw conclusions appropriate to the situation in hand.
Problem solving: Given such and such data, find X.
Perception: Analyzing, scanning environment and analyzing features and relationships between objects.
Self-driving cars are an example language understanding understanding language by following syntax and other rules similar to a human.
So now we have an understanding of AI and intelligence to bring it together a bit and solidify the concept in your mind of what AI is.
Here’s a few examples of AI: Machine learning, Computer vision, Natural language processing, Robotics, Pattern recognition and Knowledge management. There are also different types of Artificial Intelligence in terms of approach. For example, the Strong AI and Weak AI.
Strong AI is simulating the human brain by building systems that think and in the process give us an insight into how the brain works. We’re nowhere near the stage yet.
Weak AI is a system that behaves like a human but doesn’t give us an insight into how the brain works.
IBM’s deep blue, a chess-playing AI was an example. It processed millions of moves before it made any actual moves on the chess board. It doesn’t stop there though. There’s actually a new kind of middle ground between strong and weak AI. This is where a system is inspired by human reasoning but doesn’t have to stick to it.
IBM’s Watson is an example. Like humans, it reads a lot of information, recognizes patterns and builds up evidence to say, “Hey! I’m X percent confident that this is the right solution to the question that you have asked me from the information that I’ve read.”
Google’s Deep learning is similar as it mimics the structure of the human brain by using neural networks but doesn’t follow its function exactly. The system uses nodes that act as Artificial Neurons connecting information going a little bit deeper. Neural networks are actually a subset of Machine learning.
So what’s Machine learning then? Machine learning refers to algorithms that enable software to improve its performance over time as it obtains more data. This is programming by input-output examples rather than just coding.
So that this makes more sense, let’s use an example: A programmer would have no idea how to program a computer to recognize a dog but he can create a program with a form of intelligence that can learn to do so, If he gives the program enough image data in the form of dogs and let it process and learn. When you give the program an image of a new dog that it’s never seen before it would be able to tell that it’s a dog with relative ease.
Before we finish, just one last concept: Most Artificial Intelligence algorithms are Expert systems. So what’s an Expert system? The often cited definition of an expert system is as follows,
An expert system is a system that employs human knowledge in a computer to solve problems that ordinarily inquire human expertise. Basically it’s the practical application of a knowledge database. We’ve arguably only just got the first proven non expert system this year; Deepmind’s AlphaGo.
AlphaGo is not an expert system meaning that its algorithms could be used and applied to other things. Demis Hassabis, He was the co-creator of Deepmind, highlighted this in a Google Blog,
We are thrilled to have mastered go and thus achieved one of the grand challenges of AI. However the most significant aspect of all of this for us is that AlphaGo isn’t just an expert system built on hand-crafted rules. Instead it uses general machine learning techniques to figure out for itself how to win go.
Because the methods we’ve used a general-purpose, our hope is that one day they could be extended to help us address some of society’s toughest and most pressing problems from climate modeling to complex disease analysis.
In other words the algorithms the AlphaGo used to win go could serve as a basis to be applied to very complex problems. If you ought to know more about AlphaGo hit the annotation now or you can follow the link in the description.
All right, so to bring this all together and summarize all that we’ve learnt, let’s recap.
So what is AI: Commonly AI or Artificial Intelligence is a machine or a computer program that learns how to do tasks that require forms of intelligence and are usually done by humans and the other thing to take away intelligence comes in many forms and has many different aspects. At this time we just have many different types of AIs that are good a particular subsets of intelligence.
I hope that clears things up as a lot of people were confused about what AI actually is.
Thanks a lot for reading this article. Let me know how you felt or enjoyed reading it in the comment section, also tell me about any question or discussions you would like to have. Share it with the people you know and those who would like to gain a better foundational knowledge of Artificial Intelligence, also it’s history, applications and developments. Have a great day!