The most intelligent program – Artificial Intelligence

by Márton Frits, Benjámin Fehér, Moses Amadi, Fabian Tőws, Zsolt Oszlányi and Alessandro Brighenti

Can robots become self-aware? Can they become smarter than humans? Is that possible? How are they working? In this article we will explain everything in a way that even your grandmother can understand. So what are you waiting for? Go ahead! Don’t waste your time, start reading!

Many say, that in the future machines will live among us peacefully with human rights. But where are we in technology right now?

But first of all, what are the basics terms in connection with the intelligent robots? What is a computer program or an application? They are softwares, which are sets of instructions which are read by the machine […] to perform some operations (https://en.wikipedia.org/wiki/Software).

So we defined what is a program, now we must understand what an artificial intelligence is. Basically artificial intelligence is a computer program that is capable of terminating tasks, which requires human-like intelligence.

This is how Turing-Test works

It can be really simple, programmed only for basic actions, but some of them can be quite complicated as well. In the future, they may also achieve self-awareness and consciousness, and may be as smart as or even smarter than us. To measure how intelligent they are we may use a method known as the Turing-test. There are several versions of it so I will only write about the most recent one that contains three “player” a computer, a person, and an interrogator who has to tell them apart. It’s only in written questions so the interrogator can’t see them and has to be able to tell the person and the computer apart, and if he can’t do it then the computer can be considered intelligent.

Right now there’s not a single computer that could do it but even if there were then that would only mean that they can imitate human behavior and we still wouldn’t know whether they are intelligent or not.

The structure of neuronal cells

The way they make AI’s intelligent is by creating an artificial brain, similar that of a human. The human brain has cells, called neurons, which are connected to each other. In programming these cells are mathematical functions, and the connections between them are called weights. There can be multiple layers of artificial neurons, but there must be at least one input and one output neuron. The whole thing starts at the input layer. The computer gets a number. The number gets forwarded by the weights, but the weights modify its value (it depends on the value of the weight). Then the function (neuron) changes the number, and the weights forward that number and so on. However, this program must be calibrated in order to work, which means that the weights are random at first, and in the calibration process they change them.

This is how an artificial neural network looks like

AI can also learn things by itself through a process called „Deep Learning”. The computer doesn’t need to be programmed for a specific purpose at “Deep Learning”, because it’s simply fed, for example with millions of photos. This means you no longer have to write a large program, but you can let the computer learn from simple examples. This brings many advantages which should be used in the production of machinery. The machines learn independently and fix defects and errors. The machines should be connected and learn from the mistakes of others.

Computer with “Deep Learning” are also used in many other places. Such computers are used in cars and cars without drivers drove many thousand kilometers. A big advantage of autonomous systems is that the car can immediately stop in an emergency. Hence people could be saved and accidents could be prevented. Machines are already faster and stronger than we are and they always get wiser, as the example of chess robots shows because no person can match with machine intelligence.

Probably the most important usage of AI is robotics. So first of all let us define what a robot is. Although there are no fully correct definitions, there are some pretty great tries. According to one of the Oxford dictionaries, for instance, a robot is “a machine capable of carrying out a complex series of actions automatically, especially one programmable by a computer”. We can identify a robot, but cannot define them universally. As Joseph Engelberger, the “father of robotics” once said “I can’t define a robot, but I know one when I see one.”

Based on their programming, there are three types of robots: Remotely Controlled (RC) robots contain some commands that will only start to function, when somebody or something (but typically “somebody”) sends a signal to the robot remotely. Artificial Intelligence robots let the robot interact with its environment and solve problems without anything or anyone controlling it. Eventually, hybrid robots contain both RC and AI functions.

They all have three basic similarities. The first thing is some kind of mechanical construction (a “body”) that is designed just the way that they can do the task(s) they are assigned for (therefore developers have to consider in what environment the robot will be used). The second similarity is that they all need electrical components to power and control their “body”. Finally, every robot uses some level of computer programming code, the types of which I have mentioned above.

Robots can be used for many good purposes, such as tasks that are dangerous or dull for humans, or that humans can’t reach because of their limitations (like tight places or space). Fortunately, robots – like space probes, AGVs (Automated Guided Vehicles) and domestical robots (that help in the household) – will do the dirty job for us.

An example of a remotely controlled robot

However, there are also some robots the usefulness of which is questionable, such as military robots. There are robots which can move through any terrain. Such robots also help the troops to transport technology equipment. These robots can be controlled by a drone, this means you don’t have to be on the front. The drone is connected to the robot and gives it instructions, from which the robot with the help of “Deep Learning” can also learn something. So the autonomous robot is only the “longer arm” of a human. But this means that killing mass amounts of people is getting easier, which raises ethical questions. In addition, as they are getting more and more complex, the chance of them gaining self-awareness may get higher…

Robots working instead of humans are becoming more and more accurate and they cause unemployment in a rising rate. The chance of your job being done by a robot in the next twenty years seems to base on four things: how creative you have to be, how much you have to help others, whether you have to squeeze into small places or not and the requirement of negotiation in your job.

A graphic of the fifth paradigm from Moore’s Law

In many movies we see robots, which are more intelligent than humans. Since the beginning of the term ‘robot’ humans started to play with this idea. However, the big question is that will this happen in the future? And if so, when will it happen? To answer this question, we must understand what Moore’s law is. Basically it is an observation by Gordon E. Moore, and it is about the growth of speed of the computers. We can estimate, that robots will become more intelligent than us in 2045 (src= https://en.wikipedia.org/wiki/Technological_singularity). This phenomenon is also called technological singularity. But in order to achieve that, raw power is not enough, we need a good software too.

As we detailed in the article artificial intelligence – just as many other things – can be used for good, but it also can be very dangerous. So is artificial intelligence any good? Well, I think it is a question only you can answer (or see it detailed in the ethical article).