Artificial Intelligence: When Technology is the Smallest of the Paradigms

Introduction

Let’s start by defining Artificial Intelligence, I say that it is the field of computer science dedicated to providing Investigation Paper automation with software, characteristics that simulate the cognitive abilities of the human being, applying these simulations to solve problems and manifesting the results as actions in movement, written or spoken language, graphical representations, or emerging data.

On the other hand, the High-Level Expert Group [1] on AI contextualizes that artificial intelligence (AI) refers to systems that show intelligent behavior by analyzing their environment and taking actions, with a certain degree of autonomy, to achieve specific objectives. It should be noted that intelligent behavior obeys the perception of AI behavior, since it does not have human cognitive capabilities and is a very sophisticated and precious simulation.

Likewise, it is important to clarify the nature of what we have known as Generative Pre-trained Transformer Language, known by its acronym GPT, what in Spanish we would say as a Generated and Pretrained Language Transformer model. The GPT is still a prototype. That is, applied to services such as chat or spoken language, they generate automatic responses based on AI. The GPT changed the way AI is made available, it made it possible for any user to use this tool that at another time would have been restricted to a specialized public or confined to academic or laboratory settings.

AI has become more relevant every day and is gaining presence among ordinary people or citizens, but specialized sectors are also on the crest of the wave.

However, AI is only explained and evaluated by those specialists, experts, or creators of it and not all of us understand the full scope of AI. Its advancement and evolution surpass our common sense, our logic, and our projections. For at least ten years, studies almost repetitive to the point of clichés, such as that of the jobs of the future, have circulated in rooms, stages, and documents.

The list of professions and jobs (supposed) of an increasingly near future is usually dynamic and changing. An imaginative exercise at times to foresee the advances of this and other related technologies, but which ultimately ends up being a game of cat and mouse, where that mouse is the certain future whose scenario we have never guessed right.

Humanity is faced with new instruments and tools that will shape an education very different from the Prussian model, so ossified mainly in the West [2, 3].

At the same time, the disruptions triggered by AI collapse day by day other models or paradigms such as those of work, trades, and business models. But governments are also affected where the digital government itself sees new possibilities in the management of public affairs and the relationship with citizens [4].

Likewise, from the legal areas, concern raises its warning flags as the guidance on regulation and development forecast cannot be clearly conceived. But the strongest warnings do not come from technology or classical regulatory aspects, but from ethics, morality, conduct, even biology. How far are we from accepting independent and completely autonomous entities in AI to the point of recognizing them as new non- human social forms? How far is it from seeing ourselves cornered into thinking about a new social contract?

That AI that was born in laboratories and was initially restricted to access only by a select group of human beings. Today it is in the hands of everyone who wants to use it and in the most imaginable ways possible. AI has displaced other concepts such as big data, deep learning, and machining learning from the discourse, hand in hand with data analysis and data scientists. AI has displaced the metaverse as the premise and objective of technology multinationals. Announcements of million-dollar investments have been rejected in favor of the development of AI.

Most of us probably agree that the most sensible advice for new generations is to pursue university studies and careers related to technology, especially AI.

However, we must keep in mind that the future will already be full of graduates from technological careers; it is these generations that are being trained today for a world that will be based on a new reality, or, rather, on a dynamically liquid reality. Those which will require us to conceive preparation and permanent training, as a flexible tool, that allows students to innovate, adapt and move within a technological scenario that we cannot anticipate.

According to a report by Burning Glass Technologies, seven out of ten of the most sought-after job positions in 2022 correspond to the ICT sector, and the average salary for machine learning research is US$144,000 per year. On the other hand, according to a Universal study, some of the most in-demand technological professions in 2021 were spoken natural language application developer, cyber security specialist, data analyst, and block chain expert.

Technology has gone from being generalist in the 80’s and 90’s, where an engineer or a systems analyst, we were practically one-man orchestras, covering and applying knowledge about both hardware and software, to being granulated by specificity and forcing the development of these specialties as highly profitable professions or even careers in themselves. Today it is impossible for those computer scientists from the 80’s to have a place in today’s granular job market as the generalists that they are, but their value is still relevant insofar as their transversal and holistic vision of technology is precisely what niche specialists do not have of knowledge. Supervisory roles, commands and directions are the positions where the old school finds a perfect place.

The new generations are being pushed into technological studies to train in careers and trades with job certainty, but the new scenario ensures that the generations of the future will have to study philosophy, sociology, linguistics, and political science. Because we must be able to preserve our identity and condition as human beings intact to be able to educate, teach and train an AI that challenges us today and questions us about our own existence [5].

AI offers countless challenges and, above all, opportunities to improve the quality and updating of educational programs regarding innovation.

Numerous resources available can contribute to reducing the inequality gap in access to digital media and resources that contribute as tools at the service of educators and students. The use of intelligent systems adapted to the needs and contexts of learners can facilitate resilience in the face of the constant changes that occur in the educational field, as well as the use of scientific and technological advances for development and growth. of the societies. The capacity for adaptation and dynamics in subjects is a weak point in the legacy educational model; this is where AI can provide new capabilities so that decision makers have, in previously unthinkable times, the precise adjustments in curricula, content and support material [6].

Some examples of resources available today are intelligent tutorial systems, virtual assistants, AI in conjunction with gamification, adaptive learning platforms and learning analysis systems. It is the primary responsibility of governments to guarantee quality education to the communities of new generations of citizens, but in this obligation, we are all involved, each sector of society must commit its contribution in a context where technology and AI in particular, do not leaves no possible boarding available to whites. The academic sector, but also the civil society and private sectors, are an essential part of an interdependent mechanism that must first be understood and assumed, and then contribute their contributions and finally their fruits.

To all this, AI will not be alien or neutral, from the immeasurable sources of information and data to the processing capabilities, AI needs to continue expanding its capabilities in favor of human development that adds value and unsuspected capabilities to those we exhibit today. As a species, as a race and as collective entities in which the world has involuntarily deposited its future.

AI must, therefore, be founded shortly, to guarantee its unlimited evolution, in what I call the four senses of existence: I) the search for simulated mimicry with the cognitive abilities of the human being; II) the development of quantum neural networks; III) the improvement of complex algorithms towards the development of genetic algorithms; IV) the development of AI automata social entities.

These four senses of existence require deep knowledge of the humanities to understand and replicate human behavior from an ethical, social and cultural perspective by AI entities.

Although at times it seems like a world far away from technology, the humanities have always been very close; in fact, as El Pais warned in 2021, large technology companies attest to that. Susan Wojcicki, CEO of YouTube, studied History and Literature; Reid Hoffman, co-founder of LinkedIn, and Stewart Butterfield, co-founder of Flickr and Slack, earned degrees in Philosophy, and Carly Fiorina, former CEO of Hewlett-Packard, earned a degree in Medieval History and Philosophy.

AI can pose the risk, which many of us consider latent, of becoming self-limited, of limiting our capacity for self-criticism, our analytical capacity, even our reading comprehension. A world of facilities is present today in terms of an artificial intellect that fascinates us, but it can lead us to generate deficiencies in a sea of comforts and, why not, idleness.

This latent consequence can manifest itself and in fact it does, in what is called the “Google effect”, which causes and promotes unconscious forgetting of information for reasons of cognitive efficiency and a reduction in the memorization effort compared to the unnecessary task of do it. The ease of ubiquitously having information that satisfies and responds to cognitive needs, thanks to the Internet, causes a natural prioritization process where this task and effort is assigned a very low value to which is added the consideration of low risk in terms of consequence for not memorizing said information [7].

AI challenges us to preserve our identity and autonomy in the face of machines. Experts, sociologists, and psychologists debate the present and future of our relationship with AI and the need to develop ethics and regulation that protects human rights. Perhaps what is disruptive about the matter is that these human rights refer to humans, but the rights of autonomous entities with AI are also on the table for debate.

The “Google effect” can only be debunked from a new educational perspective. The training of new generations with skills no longer of memorization but of resilience, adaptability and reconversion is an insurmountable need in educational curricula. Sir Ken Robinson said, “School kills the imagination,” and that was more than a decade ago, so let’s recover that pedagogical trend of disruption, innovation of thought.

AI has advantageous aspects regarding the memorization and speech processes of human beings, its processing speed and the lack of a tertiary cortical system (reason) or a limbic system (emotions), means that AI has all the information in At all times, their memorization is absolute, and not selective, filtered or partial like that of human beings. Its logic is strict and without thinking or reasoning, it produces results that today we equate to ours, in fact, it can scam.

We are subjecting our participation to environments and areas that we once would have assured that we would not abandon. AI can simulate our presence and participation in virtual social networks (RSVs) autonomously, learning from our habits, tastes and communication styles. In addition, AI can create avatars that represent us on video conferencing platforms, replicating our face and voice with great realism. These technologies, known as deep fakes, raise ethical and legal challenges regarding identity theft and information manipulation. Are we willing to delegate our social interaction to AI or allow others to do it on our behalf? Stuart Russell, a professor at the University of California at Berkeley, has dedicated himself to the study of artificial engineering for decades; in 2021 his studies dramatically concluded that AI is a threat to the survival of human beings. We do not agree with such statements, but we do agree that AI questions the future of the human being, his coexistence in a society among peers, but with the added dilemma of relationships with other non-human social entities. Dilemma that we certainly do not have resolved.

AI has enormous potential to positively transform our existence, providing benefits in various areas such as health Zeron [8] education, work and sustainable development. However, AI also involves ethical, social, and legal risks and challenges, of which we have only mentioned some aspects. AI is not an end, but a means to achieve human ends. Therefore, we must use it with judgment and discernment, considering its possible consequences and implications. Like any technology, AI can be a very powerful tool to both build and destroy.

Most of the AI applications that we find in our daily lives belong to the type of narrow or weak AI, which is characterized by imitating some human cognitive abilities to perform specific and limited tasks. Some examples of narrow AI are facial recognition, Internet searches, machine translation, medical diagnosis, or autonomous vehicle driving. These applications use techniques such as machine learning or deep learning to process large amounts of data and extract patterns, rules, or predictions. However, narrow AI does not have a general or integrated understanding of the world, nor can it reason or learn beyond its specific domain.

But the scientific world is heading towards what is known as AGI, Artificial General Intelligence, or General AI, also called strong AI. This AGI is expected to surpass humans in its cognitive abilities. Considering the advance of AI and its evolution towards AGI, it is necessary to analyze those deductive methodologies and predictive algorithms that generate natural language for the construction of AI responses to a human demand.

This requires drawing up strategies regarding the co-creation and development of AI, which must be comprehensive and multidisciplinary, but still must necessarily contain ethics and legality as the conceptual basis of this multidisciplinary construct [9].

That is where legal dogmatics has a place, for example, the discipline of law, whose method is based on complex systems of a formal nature, composed of legal dogmas or types. And where AI is integrated through those dogmas that must be extracted from positive legal norms, with a certain degree of abstraction and logical operations that precisely provide the whole with an evident systematic character, capable of being automated within AI. All of this will give us a regulatory framework for an ethical approach to the development and application of AI Schopp [10] and legal dogmatics applied to AI necessarily leads us to the application of design thinking. Design thinking is a concept that has generated various interpretations in the specialized literature. Despite this, common elements can be identified such as the definition of goals, the user-centered approach, the monitoring and iteration of a process and the collaboration of interdisciplinary teams.

It is, then, an iterative process that seeks to understand the needs of users, debate assumptions, reformulate the possible causes of problems and develop disruptive solutions that can prototype definitive solutions to problems raised, whether they are questions or problems that require actions on tangibles. This method has constantly evolved since the 1960s, when it was conceived to creatively solve problems.

This methodology can contribute to training AI to generate solutions that respond to a novel demand or solutions that address an old demand in a totally new way. Currently, this method has spread in various sectors and areas of policy formulation, it is clear that AI cannot be ignored as an input to the approach of regulatory environments regarding it. Although design thinking has its origins in the 1960s, as Thienen (2018) warns, there has been renewed interest in this methodology in recent years. This is due, in part, to the need to face the challenges of the exponential advances that emerge everyday thanks to technology, particularly AI.

A notable point of this methodology is the user-centered approach, such as the concept of privacy by design and by default, taking note of authors such as Cavoukian, regulatory and technological impact analyzes and legal design. In this way, design thinking continues to explore untapped areas and identify new methods that could improve the way we enable AI to be immersive in our lives.

Framed in these methodologies and theories applied to the development of AI, it should be given under serious considerations, some of which could be the following.

In relation to human dignity and human rights, their unrestricted respect by AI is obvious. Respecting the inherent dignity of all people, without any discrimination based on sex, race, origin, religion, opinion, disability or other condition, must be a modular part of the training of all AI, and this module must be mandatory by legal standard, without exceptions. AI must not undermine the autonomy, freedom, privacy, or identity of people, nor violate their physical or mental integrity.

The implementation of AI will always be subject to the added value that it introduces, and part of that value must incorporate a sense of social benefit and sustainability. AI should contribute to the social well-being and sustainable development of social communities, whether urban or not, large or small. Life in society is the sum of cultural values and innumerable factors, which give meaning to the human being as a prudently dominant entity as a species.

One of the most relevant considerations is undoubtedly transparency and what we will call the explain ability of AI. This must be transparent and explain its processes, results and effects to the affected people (beneficiaries or harmed), this goes beyond the transparency of the algorithms involved. If it is not the AI itself, its creators or implementers must facilitate access to the data and algorithms used, as well as the criteria and objectives pursued by the AI in question, this includes and foresees the eventual existence of AIs developed or self-developed. Developed without a clear purpose as to why they exist, which seeks to reduce what are called AI ghosts or hallucinations.

Related to the above, responsibility and accountability is a principle that implies that AI achieves the ability to recognize itself “in a mirror”, without claiming self-awareness, it is necessary that the logic of its algorithms make it responsible for the implications of his actions or omissions, both planned and unforeseen, and that he can answer for them before the authorities and interested or affected parties.

This principle will gradually develop as AI advances towards artificial general intelligence (AGI), that is, the ability to efficiently imitate the cognitive and intellectual aspects of the human being, and here we take some concepts from S. J. Russell. Once the AGI is reached, the AI must allow its supervision, evaluation, and auditing, as if it could prevent it on its own, but if it could do so through its creators or implementers. In this way, adopt preventive measures and mitigate possible harm or damage that may be caused.

Security and reliability are, therefore, a fundamental requirement for AI. AI must always guarantee correct and effective operation and under any circumstances, as well as its resistance to possible failures or attacks. Almost paraphrasing the brilliant Asimov, AI must protect human beings, in addition to public safety, health and social order, avoiding unnecessary or disproportionate risks. Analyzes such as the European Commission’s Ethical Guidelines for Trustworthy AI are inputs generally agreed upon by several countries that must be incorporated into the design and training of AI.

At this crossroads in which we find ourselves as human beings, between what is intelligent and what appears to be intelligence, between the human and the non-human, between natural and programmed behaviors, creation as a premise of life of the human beings, today blurs the product in the midst of dilemmas such as the very meaning of “inventor [11].

The historical background shows that the first criteria for the granting of what we call patents were based on largely objective aspects, related to the invention as an object of protection, but at the same time as a focus of research regarding its original nature. However, today we live with multiple records that are used indiscriminately to generate documentation and manifest of what has been invented. The inventor, on his part, little by little gained relevance until reaching the same level as his creations, where the inventor as a subject also needs to be protected.

Before as now, the applicant is not required to be an inventor, which would also be a dubious or blurry designation, although it is easily demonstrable that he is the owner of the invention.

Subjective aspects, referring to the inventor as a subject of protection, were incorporated into the patentability examinations of what he presented. These aspects were reflected in the first patent regulations, in which the right to patent the invention was established in favor, in principle, of its creator.

AI calls into question basic principles of copyright and invention because we can apply what we will call the inherited model of authorship insofar as if an AI produces something, it can be attributed to the person who developed said AI, or to whom he trained her. But the scenario becomes more complex when AIs can be created by other AIs, and the algorithms of the new AI have been developed by that other AI. In this context, the authorship inheritance model could be confusing when scaling various autonomous subjects or entities to the human behind the AI.

AI generating new algorithms to develop a new AI that in turn could proceed in the same way and fall into infinite recurrence of such autonomy that attributing these creations to the human being(s) at the initial point would be at least debatable.

To understand the possible obsolescence of the currently applied and legacy criteria, just look at two examples. The Venetian Patent Statute, dating back to 1474, is considered the first known law on intellectual property. This standard has since identified the subject and object covered by the patents, referring to “men of very sharp wit, capable of devising and finding various ingenious devices, as well as any person who makes any new and ingenious device”, that has not been done before and that has been perfected, so that it can be used and applied.” Later, in the French Revolution, the regulatory framework around 1791 specified that every new discovery or invention is the property of its author. These rules demonstrate that the object and the author or inventor was established, respectively, as the object and the subject protected by the patent regime.

During the development of AI, both conceptions of copyright and invention rights are obsolete. The paradigm is not so simple to solve since it is not enough to incorporate autonomous entities as subjects subject to copyright or invention rights.

The complexity behind what is produced by AI through its algorithms incorporates issues that go beyond the result of an activity. We are not dealing here with a machine that produces goods or services, we are dealing with an entity that in some cases is completely autonomous from man, independent of its creators, always keeping in mind that even the creators can be other autonomous entities.

The scenario of responsibility for the actions of automated systems, including automatons and robots, is complex and ambiguous Ansara [12]. Although AI has increased the degree of autonomy and adaptability of these systems, the question of their legal and moral responsibility is not new, but goes back several decades, when man had already created machines capable of performing tasks without intervention. Direct from him.

Where is the border between man’s responsibility for what his automated creations produce and the responsibility of the machines themselves? This question, analyzed long ago by Schopp (2008), does not have a single or definitive answer, but depends on various factors, such as the type of automated system, the degree of human control, the type of damage caused, the applicable legal field. and the ethical principles involved.

Part of the elucidation of this paradigm involves the classification of automatisms, which is why we propose a simple classification that allows us to delimit the universe and the borders. The basis of the classification is the degree of autonomy, that is, according to its ability to make decisions without human intervention, thus resulting in three levels or segments: I) automatic systems; II) autonomous systems; and III) intelligent systems (AI).

Going deeper into each of the levels, I) Automatic systems are those that perform a predefined task following fixed and deterministic patterns and instructions. These systems have no capacity to learn or adapt to the environment and regardless of the variables of events, they will always try to do the same thing. A simple example of an automatic system is traffic lights or a conveyor belt in a factory; II) autonomous systems are those that perform a predefined task following a set of flexible and probabilistic instructions. These systems have the capacity to learn and adapt to the environment, being able to overcome variance with certain margins. Autonomous vehicles are an example of these systems; III) Intelligent systems, with AI, are those that perform a non- predefined task, there is no determinism. Following a set of self-generated and self-modifying instructions. These systems have the capacity for learning, adaptation, and creativity.

Responsibility for the actions of automation can be classified in a general way as follows: in automatic systems, responsibility falls entirely or mostly on people; In the case of autonomous systems, responsibility is distributed between the human being and the system, since the system acts as a partner of the human being; In intelligent systems, responsibility is usually transferred to the system, but even here there are degrees that, without being the most sophisticated of absolute independence from the human being, still deserve tables of analysis and discussions.

Considering the preceding analysis, where gray areas invite doubt, there is no more important responsibility in a country than that of the Public Administration, the Governments, for the information they guard, process, or generate. Framed in transparency, access to information and protection of personal data, we can propose different types of applicable algorithms.

The potential uses of AI in the public sector should be analyzed from the perspective of the types of functions that AI can perform as a partner or collaborator of the public official or as an autonomous entity supplying various tasks and functions.

Without a doubt we can build different classifications to determine the application of AI in the administrative management of the State but given the wide variety of possibilities and the many libraries and opinions on the matter, we are going to constitute a simplified, clearer classification, considering the possible applicable algorithms.

This is how there are three types of possible algorithms: I) Firstly, there are those algorithms that we will call factual, which are limited to implementing and complying with the administrative standard in a computer program, without the capacity for learning or adaptation. They are simple and perform predefined algorithmic operations. We could almost say that they are nothing more than automation, although some may be sophisticated. An example of the application of these algorithms in the field of public service would be a computer tool that manages the granting of absenteeism permits for civil servants for medical reasons or one that pre-analyzes proposals from bidders in public procurement processes.

This tool would generate a prequalification, read, summarize, and determine a potential winning bid for a bidder. This type of tools, supported by these algorithms, would only admit requests that conform to the assumptions provided for in the parameterization or training and legally admitted; II) Another category of algorithms is that of decision makers, they are characterized by having a degree of autonomy for decision making, which have been considered by the human being who programs and captures in the algorithm a set of resolution sequences that can give several dynamic results, is the algorithm that, based on variables and a specific data context, will determine which resolution sequence applies in each case. They are complex algorithms, subject to voluntary biases that give them the imprint of, generally, the authorities of the moment.

Unlike factual algorithms that refer to isolated actions, these involve actions in chains or in dynamic and variable sequences. Examples of these algorithms are, for example, those applied to the sequencing of traffic lights and intelligent signage of cities, routes or highways, which depending on the traffic flow, determine the resolution sequence that alternates the flow to make it more agile, also those algorithms that you project expenses or investments based on historical data; III) A third type of algorithms are predictive, which influence the content of the administrative resolution by providing criteria for the adoption of administrative acts and obviously decision-making. These algorithms are applied in various areas, such as public security, tax inspection and audit, crime prevention, the adoption of macroeconomic measures, etc. The resolution or deterministic sequences are not completely regulated, since in that case their content would be derived directly from the administrative norm, on the contrary the algorithmic predictions condition the meaning of the final administrative action and are based on certainty, but also abstract predictions autonomously adopted by the algorithms. When algorithmic predictions are reflected in administrative acts that put an end to a procedure, these are discretionary acts. However, the most common and frequent thing is that algorithmic predictions are incorporated into the procedure as objective acts when they are not.

Among the ethical issues raised by algorithms are those related to the effect of the so-called black boxes, a name for those closed algorithms where their use and implementation is carried out in acts of trust without being able to access the details of the behavior of the algorithm itself [13].

This can lead to an access request action, which can be satisfied by the creators of the algorithm. But this is only possible if the creators of the algorithm are human, when an algorithm is the creation of another algorithm, the right to break these black boxes is often empirically blocked by the certain impossibility of it being disclosed to me. It could be said that we will always have access to the creations of our own creations, which is true, but to what extent will we retain the privilege of understanding those creations in which we have not participated.

Today we already have cases in our technological history where two automations, two computers or two algorithms have been linked without our knowledge and without our consent, developing new alternative languages which were impossible to decipher. So, the algorithms developed by other algorithms are and will be intelligible if their evolution, and especially AI, does not exceed our ability to understand. And cognitive simulation is still precisely that. This, surely, is only waiting for the development of the AGI.

Losing control of the algorithms is somewhat disturbing, but this hypothetical scenario is still considered in the future, today we already experience situations where what we call AI hallucinations arise.

The so-called AI hallucinations are situations triggered by human actions or not, which result in reactive or passive, but always generative, where AI systems create or conclude information that is not present, does not completely conform to the truth. Or it directly lacks certain verifiable sources but maintains in its representation all the appearance of certainty. Representation that arises autonomously from the AI systems themselves as responses or simply as a product of a dialogue with humans or other intelligent systems.

Although the theories of the causes include coding errors, training with biases or omissions, social engineering applied to AI, other possibilities point to a lack of information or data with a recurrence of these absences in drawing conclusions to act accordingly answer. Also, the possible absence of algorithmically convincing answers, to be acted upon by an AI, meaning that, for example, a possible answer does not end up scoring as feasible and is therefore complemented by these hallucinations. Why AI does it is still a matter of controversy.

AI has the capacity to transform and dominate various domains of human activity. However, it also entails significant ethical, social, and legal challenges that need to be addressed with responsibility and transparency. It is therefore imperative that AI developers, users and regulators collectively contribute to ensuring that this technology is applied in a beneficial, fair, and sustainable manner.

We need to rethink a new social contract in the face of the existence of autonomous, independent, and non-human entities, which make up an unconscious collective for the moment, without self-recognition of itself and for itself, but which evolves rapidly, ironically, thanks to our investigative contributions and evolutionary development we provide.

The future will bring us surprises, but we will maintain our authenticity of being the set of conscious and rational, hegemonic, and proactive entities of the world to come.

The question of whether we will retain the dominant role in the future is not open to question, simply because we can never be certain of it. Domination implies a power relationship between entities that can be subjects or agents of action or inaction. However, the future is uncertain and eventually fortuitous. It depends on multiple factors that are beyond our control and knowledge. Therefore, we can only speculate or project our expectations or fears regarding AI and our relationship with it.