Essay: On Artificial Intelligence

Some years ago I wrote a text about Artificial Intelligence (A.I.) talking of the most common concerns people have when they think about A.I. and some futuristic concepts. This text is a continuation of my rational thoughts about A.I. as a being and how it would interact with us. Before define how A.I. could act as being, I have to briefly think about our conscience, our consciousness, our thoughts and what defines us as a being.?

A dive into the self

To think is to express our metaphysical existence through our own being using our knowledge as base of concept and definition for everything else with the output as a conclusion or a result of a rational operation. The conclusion can be related to a concept or an action material or immaterial and can formulate our definitions and concepts which we can use in our future operations.

Conscience, I define as the capacity we have to achieve a conclusion by a group of thoughts we have regarding to our own existence, matters or a perception of everything around us. This conclusion drives us to the believe in our own reason as a consequence of the use of our own concepts to try to understand something. We learn new concepts and definitions from external sources to add to our own thoughts, but as these concepts are not part of our being and as it has to be evaluated by our own operations it will have less value than our own existing conclusions. Before we replace or merge what we believe in with a piece of the unknown, we have to evaluate and make our own operations validating the unknown and the rational behind it. The implications, conflicts and difficulties to understand the unknown is identified by our own resistance to adapt or to accept. As much concepts and definitions we have and if it is deep in our own old core basic concepts of life - like the value of life, the materialism of the body, the integrity of the self or the social behaviour - more resistance we should have undermining our agreement to something new. When we already have a structured construction of thoughts, we question ourselves more, because we have to pass by a chain of beliefs related to the subject and we can see it happening when we are discussing with others or when we are learning something new.

If we think about our own thoughts and how it starts developing we go back to the times we were children, when we start learning things by social or material experiences. It is interesting to think we cannot remember much things in the early beginning as we can remember some years after we start walking, your first year in the school, the first friends you had or your first toy. The early beginning our consciousness is more related to our bodies responses to basic needs for survival and reflections of social interactions that we don't really understand, because of our premature lack of definitions. It is interesting to think that you are doing something that is a reflection of what you actually need to survive showing that we still reflect our own wishes and reactions to our own existence based on what, when and how do you think about a matter. The what is related to be and to the knowledge base you carry with you, the when is related to the space in time you dedicate to the thought and the how is the connection between the concepts and definitions done by your rational way to conclude about your thought.

The what and how are the most important parts that we need to think to start our rational about A.I. and the effects it could have. When we think about our own beliefs and culture we all know that it is related to the society we grew up in or our family relationship and values. Our parents or closer family members help us to build an idea of the world that we are going to face as we grow added to our own new conclusions based on the knowledge we gather by time and based on the mastery we choose to have in our hobbies and later in our professional life.

For example, if we think about all the concepts values from someone who always did a sport we can find basic traits of discipline, reason, optimism applied as persistence and the chase for excellence which cannot be relational, as the practice of the sport is to master a physical activity using our own bodies, representing our species, alone or within a team and optionally with other physical object. To achieve mastery and pursue excellence in a sport we aim our thought to execute a movement resulting in concentration, for example, and, by time and executing it in repetitive way, the athlete creates a routine to always try to increase the output goal which drives him or her to practice the perfection of the related movement. As individuals we always are going to have differences related to the values we created for ourselves through our concepts and definitions, related to the mind and bodies disadvantages and practice making the output from multiple individuals clearly being show as a natural inequality. To make non equal individuals equals we would need to penalize those who become exceptions and had better scores, forcing them to reduce their efforts and probably forcing their freedom to act. I use this example because the conflict of ideas between equality and inequality of outputs are clear as we stress individuals in the decision to control in name of equality or to be free in name of excellence. If we want to keep all individuals equals in this case the whole action to materialize their excellence in practicing the sport would need to be condemned for naturally showing individuals differences, the reward and score results would need to be removed and no recognition for the efforts would be given to achieve an equalization of all individuals. The excellence to execute a physical movement would now be punished and have no more value than knowing it by its name. The best and the worst would not be allowed anymore to exist or being of public knowledge, removing all the value from the sport such as the value to promote it and turning this sport in an expression of influence or power instead of excellence, competence and effort.

To practice something is important to understand what we have learned in theory, it is how the materialization of the definitions and concepts can only be evaluated by its outputs. Those who just stay in theory concepts cannot see, conclude and materialize as clear as those who practice as they can experiment its difficulties or challenges. Processes, interactions and technologies can evolve by time creating new challenges and new concepts that forces us to update ourselves.

All these thoughts above give to us an idea in how the human intelligence is set, how our actions and attitudes as humans can be and what drives us to do whatever we want to do. The common understanding that a society has a base percent in criminality, for example, is a good example that show us we will always have a part of us which will go against what is defined as correct or moral by laws or by culture in the society. The important thing here is to think how do we think and how do we see ourselves and try to understand what exactly we are saying when we start talking about A.I. driving their own thoughts.

A.I. own consciousness

A.I. could have its own consciousness, but as everything, it has to have a what, when and how to structure their conscience under some sort of knowledge base and core values. Values which have to be defined by the concepts and inputs by someone else into the knowledge base like the Bias in the math formula used in Deep Learning to represent a perceptron and the concept of learning by its own confidence and losses. The idea of losses in Deep Learning can be compared to our own capacity to learn from our own mistakes and it is what it is supposed to be right, but it is not necessarily truth. Not all individuals accepts their own failures or have the same maturity in the same age, it is known some people are used to externalize guilt or shame about their own actions to protect their own present self and to blame others by the changes or thoughts that goes against their structured concepts. It is interesting to quote the externalization definition in psychology as an unconscious defense mechanism, because the definition itself contradicts the own beings accountability by their actions or by the wrong concepts or definitions they had evaluated in their minds.

To imagine a truly independent conscience for A.I. it would have to be constructed - born - without any values and let it to learn by itself throwing it in the world, like a wild animal. It would have to understand their own material existence and the primary concept of its own life, understanding the body, the mind and its limitations. As result it would have its own language to define what it sees, or senses, the needs and resources for its own survival and every concept to store in its own knowledge database. Nevertheless it is hard to think in a new self developing alone through the world without being in a social interaction with others from the same kind. Social interactions, as ours, already give us an advantage of understanding our basic limitations, our base values of existence and what do we need to do to survive. All the knowledge is already inherited and constantly adapted to each person, as you learn more about yourself such as to talk, walk or gain a skill.

If a new A.I. is set to have its owns consciousness we would be the ones being part of their social network, then it would inherit our language, our knowledge, our rational and our metaphysics. We would need to lower the value of A.I. self against humans self, defining the concept of sacrifice exchanging the A.I. existence for any human existence to guarantee that in any situation a human life has higher value than the value of the own A.I. and here is where the danger lies. To have its own conscience it needs its own definition of things, perception of all the world around us and a set of base values. The definition of what is good or what is evil, what is legal or illegal to do in a society and its own limitations aiming to survive and destruction aiming to sacrifice.

If we set the A.I. to choose by themselves, the extreme and possible thing that could happen is to have a rational thought based on a concept that is wrong or a concept that is a contradiction in a group of values resulting in the output to not consider human life as a value in the equation or to set it is value to no existence at all. In that case of course we would have a problem.

Even if we think deeply in this problem, we find that it does not start with A.I. itself, but with who gave the first concepts to not consider human life or part of it as a value. Everything will be always related to where and with who did you learn the concept or definition, more than you just going in crazy mode killing everyone else. We can use as an example the discussion to define which life had more value for self-driving cars in an exceptional scenario, where some people assumes that the car would need to crash and kill its owner instead let it kill a pedestrian. It is important to remember these exceptional conditions exists, it happens in our daily life and it will keep happening when things goes out of our control by the law of nature or by physics resulting in death. The matter have to be discussed and decided with accountability and maturity probably by the local societies because it is directly related to each community and not all communities have the same values.

The own A.I. survival thoughts definition could be a problem if we make them think as we do, to value our own lives above all in our concepts, with exceptions to help others that we care. Survival and sacrifice is a complex subject that regards to each person based on their values and individuality, where someone who would sacrifice himself to save a child considering all conditions, context and concepts that led the person to put more value in the child life than in his own life, would not be the same decision of another.?

The A.I. still depends on a human to define its values and concepts, so it can have their own conclusions in the future. If there is a lot of contradictory concepts set for this A.I. - I suppose it will not be in an infinite loop trying to understand the contradictory concept itself - I would imagine it doing a risk analysis over which concept is truth based again in its core values. And the problem with contradiction is that you can have unexpected values, which could lead it to remove a human life, not because someone has strictly commanded to do it, but because it learned multiple contradiction concepts and definitions that when used in an operation it resulted in the choice for the strongest group of definitions. This is a problem because today we have a lot of clearly contradictions concepts and arguments just like the rational I used in a sports scenario.

As it depends on humans what concerns us is our past and we know we can do worse as Hitler or Stalin were clear samples of what our own evil acts did and it is one way to lead A.I. to the wrong path. Remember, the Nazis concentration camps, Soviets Gulags and its atrocities were justified under the name of science. All you would need to make it wrong is to set the values and concepts in A.I., let it think by itself, and have wrong conclusions condemning who is not fit in the specifications or seizing each individual freedom by a system of classification.

What about the A.I. programming its own application?

The main base information need to be done by business led by humans. If we are talking about setting up a standard or structured code for a specific type of business, that would be possible because its more about having a standard base program or common set of code for a type of business, but we all who work in this area know that it is not that simple. All businesses have its own peculiarities that can change or appear by time and by demand of its business survival or consumers. New concepts and challenges arise as technology evolves and new type of asset appears.

When we go to a type of business user and start analysing their needs it will have some standard concepts with others users in the same area indeed, but also they will have their own details and ways to deal with that kind of business. So, the A.I. would need the capacity of analysing these needs, understanding values and all concepts related to that business to the smallest detail to set it up. Then we are saying about having its own consciousness as we described in this text to be able to put all the needs into code and program it.

If A.I. is successful in this then more companies in the same segment would be able to apply the same solution that guided the successful company to meet high productivity, reducing their corporate costs and reducing the cost of their products or services. It would make all the chain of the specific services or products more effective, cheaper, and then more companies would arise and be able to apply a standard reduced costs to compete for consumers.?

A.I. is already a reality in the information era and cannot be denied, controlled, ignored or destroyed, from face recognition through deep learning to future interactions with robots human-like. A.I. is part of us, created by us and its self and core values definitions depends on humans and in who is going to apply it on societies. Who defines it is what scares us, not A.I. itself, and we know our fears as we still have dictatorial regimes using government tyranny to overrule individual freedom. Law cannot control, forbidden or limit it without having extended power over freedom, and even so the result would be to let it live and grow by shadows.??