Learn about: Artificial Intelligence and Machine Learning
Artificial Intelligence is one of the most important topics of the present and future. We call machine intelligence artificial intelligence. And intelligence is the ability to acquire knowledge and apply it. Extraction of machine knowledge is machine learning. The two together mean learning and applying artificial intelligence. Artificial Intelligence and Machine Learning.
Robots we have seen in movies or science fiction become very powerful. Destroys people. Some robots work to benefit humans. The brain of these robots is artificial intelligence. Without artificial intelligence, robots are like ordinary inanimate objects. Not that we need robots for artificial intelligence. The smartphone in our pocket has many apps with artificial intelligence.
Steps of Artificial Intelligence:
Artificial intelligence programs can be divided into three phases.
ANI or Artificial Narrow Intelligence or Weak AI. ANI programs specialize in a specific task. For example, a program that can play chess can play chess best. If you let him do something else, he can do nothing. Like if he is allowed to play Ludo, he cannot. The program will look at you the same way young children look at you when you ask them to do something difficult.
AGI or Artificial General Intelligence: The second stage of artificial intelligence is AGI. It is also called Strong AI, or Human-Level AI. At this stage, the computer will be able to think like a human, plan like a human, solve problems, suddenly look around in a new environment and get the ability to adapt to that environment.
ASI or Artificial Super Intelligence: The third stage of artificial intelligence. When the computer becomes more intelligent than humans, we call it Artificial Super Intelligence. Researchers are now worried about what will happen if an artificial superintelligence system is created, whether people will be good or bad.
Scientists are now very worried about whether we will survive after reaching Artificial Super Intelligence. Whether we live or not, reaching ASI will change everything. How to use all the energy in the world, we can figure it out. Maybe we can get out of the earth and go to other planets easily. I can roam the universe.
What stage are we at?
We are still in the Artificial Narrow Intelligence phase. All the programs we see or use are ANI programs. Google Search Engine, Self Driving Car, Flight Control, Nuclear Project Control, and other complex systems are all ANI programs. They are skilled in a specific task. I can’t do any other work outside of that work.
In 1997, IBM’s Deep Blue computer program defeated a chess grandmaster. Which was a smart program. It was a great achievement of artificial intelligence.
In March this year, the AlphaGo program defeated the world champion of the Go board game. It is a much bigger achievement than the first achievement. Because in chess, there are about 25 possible moves from each position to the next move. 200 moves are given in the Go game. As easy as humans can figure out which move to give, it’s not so easy for computers. Besides, chess is a logical game. Can be played with logic. But the game of Go has to be played with intelligence. You have to think the way people think.
Also, in 2011, IBM Watson defeated two champions of America’s famous quiz show Jeopardy.
IBM Watson is the most successful artificial intelligence program ever. It was initially developed to answer questions. Jeopardy’s journey began by defeating two born champions. Now IBM Watson can be used in many places. This program can be used in different areas like data analysis, education, medicine, etc. IBM Watson helps doctors make decisions about cancer treatment. In addition, nearly 90% of lung cancer nurses work based on decisions made by IBM Watson. The best part about this program is that the more it is used, the better the system will continue to improve.
Anyone can build intelligent software using the Watson API. Not free though!
Personal assistant sounds like a science fiction story to us but now many tasks can be done with a personal assistant. He will answer anything you ask him. You can communicate with the computer by talking with products like Amazon Echo or Google Home. As seen in science fiction movies. Although there were stories before, things are now real.
The robots that we see like humans in movies are called humanoid robots. The best humanoid robot so far is Honda’s ASIMO robot. Besides, Boston Dynamic is working on military-grade robots. Their robots are very strong. NAO is a humanoid robot open to everyone. It is programmable. You can buy it and modify it to your liking. Setting programs like IBM Watson on it will work like Google Home or Amazon Echo.
The most interesting robot is the self-assembling robot. For those who have seen the movie Big Hero, Microbot is shown there. Small robots. Which together can take any shape. If you watch the movie a few days ago, you can see that such a robot is actually being worked on. And succeeded. MIT’s Self-Assembly Lab is working on such robots. And made great microbots.
When nanotechnology is advanced, it will be possible to cure any disease of the human body with these self-assembled robots. Viruses in the body? The microbot will enter the blood cells and destroy the viruses. There will be no disease like cancer.
Not that the robots will look like humans. Robots can be anything. Many jobs in the industry are done by robots. From making cars to making mobile chips or processors, jobs are done with robots.
We are solving many problems just by using ANI. Made our life much easier. We do not directly tell you what to do when creating these ANI programs. We give some examples. Then the program builds an algorithm from that example. After that, the next thing that is asked to do is do it in your own way. This is machine learning.
All the programs or robots I talked about a while ago, of them have used machine learning. Machine learning will be used in all software or apps in the next two to three years.
Many of us have used the Prisma app. For those who haven’t used it, at least I’ve seen pictures made with Prisma. What makes an ordinary picture a beautiful artistic picture? This app is based on Deep Neural Network. Which is a method of machine learning. Now we have to apply machine learning to do great things.
Machine Learning Job Market:
UpWork shares the Skill Index every four months. Now the most growing skill is machine learning! Not only in the future, but there is also a lot of demand for machine learning in the present as well.
Learn from where?
There are many resources online. My favorite site is Udacity. Apart from free resources, there are also premium resources. There are two projects on GitHub called Awesome Artificial Intelligence and Awesome Machine Learning. Here is a list of all the resources on artificial intelligence and machine learning. You can start with what feels easy to you.
Another type of neural net has been researched since the 1990s, called the convolutional neural network, or CNN for short. Its application is the analysis of spoken words, images, videos, or in general data streams from any sensor. A characteristic of this type of data is that two pieces of data near them are close in space or time and are therefore closely related. For example, a spoken word is joined together to form a meaningful word, or an object can be recognized by matching several adjacent dots in a picture. In a traditional layered neural net, each neuron in a layer is connected to each neuron in the layer above it with appropriate weights. Not so on CNN. Some neurons near one layer are connected to a neuron in the next layer. This pattern is repeated in all connections between these two levels.
Not only this, the sum of these connection weights is kept the same for all connections between the two layers. The reason for this arrangement is that it doesn’t matter if a pattern – be it a particular word or image – is slightly different in time or space. As it is particularly effective for pattern recognition, it requires very few connection weights to be set. Hence its learning is done much more easily than two fully connected levels. This is called the convolution layer.
In a convolutional neural network, the convolution layer is followed by a pooling layer. Among the several neurons near the convolution layer, the one with the highest activity, its output goes directly to a neuron in the pooling layer. Normally the number of neurons in the pooling layer is very small. From this small number of neurons, a detailed image or information stream is extracted or extracted.
By successively several such convolution and pooling layers, a concise description of the image or data flow is obtained. This entire network is taught with many examples using a back propagation algorithm, adding a few fully connected layers at the end. Thus, combining the special structure of CNN and the ability to learn complex relationships of multi-layered nets, the net created was demonstrated in the 2012 ImageNet competition. In this competition, CNN is clearly superior to competing algorithms . Since then, CNN has become indispensable in image recognition and understanding. It also excels in understanding spoken or written words.