What is AI. AI and Machine Learning.
Blockchain, artificial intelligence and machine learning have been extremely popular trends in the last several years. Often, people would use terms “artificial intelligence” and “machine learning” interchangeably, even though these expressions mean very different things. This series of articles would first explore the concept of “artificial intelligence,” then look closer at machine learning and finally discuss how and why blockchain, artificial intelligence and machine learning could work really well together in one system.
One of the issues with the concept of artificial intelligence is that the understanding of the concept has been changing as people were inventing new technologies.
Artificial intelligence throughout the history
The idea of artificial intelligence is present even in the myths of the ancient Greece with blacksmith god Hephaestus creating a giant bronze mechanical man Talos and tasking Talos with protecting the island of Crete from pirate invasions. Talos had just one vein, but the blood in the vein was divine and came from the Olympians, who were the twelve main gods that ancient Greeks believed in.
When people have first invented the calculator, they viewed the calculator as a “logical machine” that had some capabilities of a mechanical brain. Eventually, computers became capable of performing extremely complex mathematical computations.
However, the issue with these computations is that the computations follow sets of precise rules, which is not what engineers consider to be “smart” in the twenty-first century. For instance, when you scroll down a screen, or when you move a cursor on the screen of your laptop or desktop, your computer simply follows exact instructions that it has in its code. The machine does not make any decisions in a way that a human would make them.
An example of a task that requires artificial intelligence
Artificial intelligence in the twenty first century is more about accomplishing tasks without being told by humans how to do so. For example, even if you feed a machine tens of thousands of pictures of nature, it would not be able to do anything “intelligent” with these pictures.
Let’s say that you want to know how many pictures out of thousands of pictures that you have uploaded contain lions. To do so, you would need to write code that first defines what a lion is. If you don’t do that, the machine simply wouldn’t know. It could store all the data, but it would need instructions before anything happens to the data. Once you define a lion, you would need to give the machine instructions in the form of “if, then… or else” code. All of this is not what is considered to be “smart” in the twenty-first century.
Obviously, having machines perform operations to accomplish the task from the previous paragraph imposes a lot of limitations on what a machine can do, which significantly limits the scope of applications. One of the biggest limitations is that in many cases to have a machine do something, humans need to spend a large amount of time writing code, tweaking code, de-bugging code and making sure that there are no mistakes in the instructions that they create for the machines. It is humans and human work that are currently a bottleneck for what machines can and cannot do.
While algorithms and software similar to this example, when a machine needs to find something in a picture, do exist and have been created in the past, such software is not very useful if there is a more advanced task, say, you want a machine to monitor a live video stream and send a notification every time a lion enters the frame.
Artificial intelligence vs machine learning
Today, artificial intelligence is about having machines make decisions similarly to how humans make decisions. One of the issues with this and with the definition of artificial intelligence is that people still know very little about how the brain functions and how humans make decisions.
Scientists do agree that the brain is the center that sends commands to the entire nervous system of a human body. However, when it comes to questions such as “what exactly is consciousness,” “what is the connection between brain and personality,” “why do people have dreams when they sleep and what happens in the dreams,” there are simply no answers that the scientific community agrees on. What is even more important from the perspective of working with machines and artificial intelligence is that scientists still do not know how exactly our brains process and store information. There is no specific data about where and how it exists, how the brain accesses the memory, and so on.
Machine learning is a subset of artificial intelligence and is about teaching computers to use data in a similar ways humans use data when making decisions. This means that machine learning is about teaching machines what to do with data and also about allowing machines modify what they do as the data or the volume of data change.