Artificial Intelligence vs Machine Learning
For dummies - In memory of Margarita, my grandmother!
What about Artificial Intelligence (I.A.)?
Throughout the history of humanity, there has always been the duality between dominated that serves me and servant that is revealed. This duality in thought has given rise to all kinds of elucubrations based on fear and becomes the engine of stories of all kinds, including wars.
This type of thinking was taken from the imagination to the machines that flourished in centuries before the 20th, in the industrial revolution, and which were considered with the potential of being faithful servants, but that by adding functions that allowed them to exhibit some intelligent behavior they were associated with the old fear of the servant who reveals himself.
This is how in recent times all kinds of theories of how intelligent machines take over the world have emerged in us as a living species on planet earth. To this can be added the fuel of the resurgence of Artificial Intelligence as a discipline that invades and improves each of our daily activities.
A Czechoslovak work published in 1917 by Karel Kapek, called Rossum’s Universal Robots, gave rise to the term robot. ‘Robota’ is a word from Czech that means servitude or forced labor, and when comes into English it mutes to the term robot. This narration tells about a brilliant scientist named Rossum and his son, who develops a chemical that is similar to protoplasm. They use this substance to make robots, and their plans are for robots to serve the human class obediently to perform all physical work. Rossum continues working and makes some improvements in the design of robots, eliminating unnecessary organs and other elements, and at the end develops a "perfect" being. The plot experiences an unpleasant turn when the perfect robots begin not to fulfill their role as servers and rebel against their owners, destroying all human life.
For 1984 "The Terminator" is released, a science fiction and action film. History tells us that, in 2029, after devastating the Earth and enslaving humanity, the machines, ruled by an artificial intelligence known as Skynet, are about to lose the war against human resistance led by John Connor. Faced with this situation, the machines understand that killing John Connor in the present would be irrelevant, given that it has already led the human resistance of the entire world to victory. Therefore, Skynet develops its strategy by deciding to eliminate the enemy leader before he is born, so that his future driving mission cannot be fulfilled. To do this, he sends a Terminator T-800 model Cyberdyne 101, a cyborg killer, through a time machine, with the mission to exterminate Sarah Connor, John's mother, before it is conceived.
Running the year 1999 a cult film is introduced: "The Matrix". Here, Thomas A. Anderson is a day computer programmer and a hacker named Neo at night. He has been intuiting all his life that there is something else, that there is something that fails and that doubt is reaffirmed with a message received on his computer: ?Matrix owns you?. Thus, Neo begins the desperate search for a person he has only heard of: another hacker named Morpheus, someone who can give him the answer to the questions he pursues: what is Matrix? and why do you own it? Morpheus and his team, realizing that his enemies are looking for Neo, decide to get in touch with him. Hacker Trinity, Morpheus' friend, leads him to him and the answer he seeks. But to get it you must give up your previous life and everything you had known before. The symbol of this process is to accept taking a red pill; instead, the blue pill could return it to its current world without, apparently, nothing that is happening would have happened. Neo agrees to take the red pill, forget his life and everything he knows to discover "what is Matrix."
Neo discovers that the world he thought he lived in is nothing more than a virtual simulation to which he is connected by a cable plugged into his brain. The billions of people who live (connected) around them are being cultivated in the same way to be able to power the machines. This collective illusion (or interactive simulation) is known as the Matrix (the matrix).
The group of real-world rebels led by Morpheus rescues Neo from the harvest of people where he was imprisoned. Once released, Morpheus explains what reality consists of: because reality is not reality. They are close to the year 2199 and humanity is enslaved by machines, which after the development of AI (Artificial Intelligence) rebelled against its creator, man. The revolution led to a great war for the survival of both contestants. This, in turn, led to the deterioration of the environment making it unsustainable for man and machine.
The common denominator in all stories about artificial intelligence is the fear of the servant who reveals himself (The machines), due to his self-consciousness and emotions are exacerbated by presenting that they also extinguish us.
But what is Artificial Intelligence (I.A.) specifically?
When a machine imitates the "cognitive" functions that humans associate with other human minds, such as "perceiving," "reasoning," "learning," and "solving problems" then the term artificial intelligence is applied. As machines become increasingly capable, the technology that was once thought to require intelligence is removed from the definition. For example, optical character recognition is no longer perceived as an example of "artificial intelligence", becoming a common technology, but today technological advances still classified as artificial intelligence are autonomous driving systems, those capable of playing Chess or Go and automating tagging on Facebook. This shows us that the term is dynamic and associated with what may surprise us.
Well, and how can a machine become intelligent?
The first thing to be clear about is that machines are only capable of processing data, but since the machines are programmable, it is possible to build higher levels of complexity with the data they process. Let's look at the following:
[ Source: https://twitter.com/gapingvoid ]
From the previous graph we can present as examples at the data level: "Gonzalo", 21, "Road", "34", "Gomez", "66", "Cali", "Colombia", "Cat". All of them are data. But with this alone it is very difficult to build intelligence because if I am a Spanish speaker I know that "Gonzalo" is a name, but "Victoria" is ambiguous for identification.
From the above, we can understand that complementary data is needed that defines data (metadata), and allow us to obtain information through that definition. Thus, if I have "First Name: Gonzalo", "Sure Name: Gomez", "Address: 34-66 road 21", "City: Cali (Colombia)" I can generate a meaning relevant and intended for the data.
On a next level, we can talk about the knowledge that is where you can build definitions from the information with the possibility of building generalizations, such as that a person can have "First Name" and "Last Name", live in an "Address" located in a "City".
Then I can have revelations through the identification of non-trivial relationships between different knowledge through intuition. Here, you can talk about inspiration or innovation. Thus, we could find that the people of “Cali” love the “Cat”.
Finally, we are talking about creating wisdom through the awareness of the connection between knowledge of knowledge. So, we might discover that it is what makes people in “Cali” like “Cat”.
The challenge for the machine has been to have adequate programming and sufficient processing capacity to be able to process the data in aggregation and association sufficient to produce intelligent results. Without sufficient capacity, it is not possible to reach the data processing that allows obtaining wisdom. And that is what is being achieved today, hence the sound of the theme. The limitation that persists is that although the machines already have intelligent behaviors, they still do not have self-awareness, which incidentally is a great concern for humans.
And, they can extinguish us?
As we have seen the machines produce a result according to the developer's will that the program, so, the machine only has and will have the autonomy that a human assigned. This is known as the "human bias," which is the bias that the human introduces to the machine at the time of programming.
And, the machine learning?
Machine learning generally is to understand the data that will be processed and adjust models than resolve the problem. Algorithms are sets of explicitly programmed instruction in traditional computing used by computers to calculate or solve a problem. The algorithms of machine learning allow the computers to train on data inputs and use statistical analysis to output values that fall within a specific range.
This means that I train the machine as a mammal trains its children. Training by example instead of words!
We can have different algorithms of machine learning:
Supervised Learning
[ source: https://miro.medium.com/max/908/1*toRnJNryryhKETfBqOcvVg.png ]
If I need to classify fruits the way is to show the fruits to the system for learning and next validate its learning to adjust what is learned. For that, we use labels for each fruit in the training process. After that, I can process new fruits and the system will classify it automatically. In the case of a new kind of fruit, I can re-train the model.
The clue is understanding what are the attributes that can let identify accuracy each fruit like color, size, shape, height, etc.
Unsupervised Learning
[ https://qph.fs.quoracdn.net/main-qimg-e510d5175c56d0b7d78e8c59a7a8c8d5 ]
In this type of machine learning, we do not have the labels available for our samples. The chalenge for the machine is to group and identify information according to similarities, patterns, and differences without any prior training of data.
Reinforcement Learning
In this type of machine learning, we have two agents. The first one executes an action to the environment and the second one interprets the results and gives and state and a reward.
Let’s see:
The agent input is in the way as an initial state from which the model starts. Next, there are several actions that can be performed at a specific state which leads to different states with different rewards. It is a goal for an agent to maximize the reward function.
Deep learning
Deep learning has as a goal to imitate how the human brain can process light and sound stimuli into vision and hearing. The biological neural network inspires deep learning architecture and consists of multiple layers in an artificial neural network made up of software, CPU, and GPUs.
[ Deep Learning = Learning Hierarchical Representations Slide by Yann LeCun, all rights reserved. ]
I know you are in heaven, I hope you enjoy this document with a great cup of coffee (Ginebras Coffee, your favorite)
References
https://machinelearningmastery.com/what-is-deep-learning/
https://www.digitalocean.com/community/tutorials/an-introduction-to-machine-learning
https://blogs.sas.com/content/subconsciousmusings/2017/04/12/machine-learning-algorithm-use/