Frequently Asked Questions
Does artificial intelligence truly think, or is it merely an illusion resulting from language and tools?
The impression that artificial intelligence thinks results from the use of linguistic metaphors and anthropomorphizing cognitive shortcuts. In reality, these systems are elements of a process of externalizing cognitive operations, which rely on mechanisms of cooperation and the accumulated work of humans designing data and infrastructure.
1. Does AI actually replace humans, and in what way does it "think" compared to the human mind?
2. AI does not replace humans as an autonomous subject; rather, it is a tool implemented by people and dependent on the work of "invisible personnel" (data labourers). Language models are not simple databases but learn statistical regularities, which allows them to reproduce some human conceptual representations, although they lack consciousness and full sensory and motor grounding.
3. Why does AI seem more intelligent to us than other complex technologies?
4. AI seems more intelligent than other technologies because it has crossed a rhetorical threshold by entering the domain of speech, which is recognized as evidence of a person. Unlike other complex systems, language models operate in the space of subjectivity and are capable of conducting a dialogue.
5. Why do we have the impression that AI thinks like a human, and what actually lies behind this illusion?
6. The impression that AI thinks like a human results from the fact that it uses sensibly sounding utterances, which reflexively prompts us to look for an intellect on the other side. In reality, behind this illusion lies an extensive collective system comprising mathematical models, technical infrastructure, and vast linguistic corpora that are the result of the work of many people and institutions.
7. Did the idea of automatically generating answers from symbols exist before the invention of computers?
8. Yes, this idea existed in the form of so-called "letter magic" and dreams of automating inference long before the creation of computers. Examples include medieval systems such as zairadja or the wheels of Ramon Llull, which used symbol manipulation and rules of combinatorics to turn questions into answers.
9. Why do AI systems generate answers that sound plausible but are untrue?
10. AI systems rely on combinatorial procedures and the statistical probability of phrases occurring, rather than on access to reality. The internal regularity of the sign system and the correctness of operations on representations do not guarantee that the generated answer is consistent with actual facts.
Do modern AI systems and ancient logic machines actually understand the concepts they operate with?
These systems do not need to know concepts experientially; rather, they operate on symbols assigned to them according to the rules of the system. This creates a distinction between lived semantics and operational semantics, where the ability to correctly use a representation of a concept may be merely a syntactic-relational proficiency.
Are AI systems objective, and how do modern language models relate to earlier attempts to create a universal language of reasoning?
Algorithms are not objective because every generative system is based on a specific order of values and classification (e.g., datasets or objective functions). Modern language models continue a centuries-old intellectual problem, initiated by figures such as Lull and Leibniz, concerning the possibility of decomposing reasoning into elementary symbolic operations.
Can the ability to obtain a correct answer be separated from the necessity of understanding the process that leads to it?
Yes, competence can be partially separated from the competent person and transferred into a procedure, an artifact, or an institution. This allows for the use of tools that provide a correct result without the need to independently reproduce the entire intellectual process leading to that outcome.
How do Baroque calculating devices relate to modern theories of intelligence, and does the automation of tools affect our cognitive abilities?
Baroque computing devices are examples of distributed cognitive systems, in which the thought process occurs in the relationship between a human and an artifact. The automation of tools allows for the offloading of memory and attention; however, it may lead to a weakening of internal cognitive abilities, as external proficiency does not guarantee understanding or wisdom.
Does a system's ability to generate correct answers mean that the system actually understands the content of those answers?
Not necessarily, because the correct manipulation of syntax does not have to imply a semantic understanding of the content. However, there is a philosophical dispute in which some researchers argue that semantic properties can be possessed by the entire system or a system coupled with its environment.
How do early projects of universal languages and classification systems relate to the functioning of today's AI algorithms?
Early projects of universal languages and classification systems, such as Wilkins', aimed to create a precise taxonomy of the world that eliminated ambiguity. Modern AI algorithms rely on a similar mechanism, as machine learning requires labels and categories that are not neutral copies of nature, but rather reflect prior ontological and political decisions.
Does the automation of thought processes in AI inevitably lead to the loss of human competencies?
Automation does not lead to the loss of competence in an absolute way; rather, it depends on whether the tool replaces a low-level activity or an operation that is the goal of learning. However, there is a risk of so-called automation irony, where over-reliance on reliable systems makes the user more efficient at performing procedures while simultaneously losing their understanding of the meaning and causal competencies.
How does the Baroque approach to gathering knowledge differ from the nineteenth-century concept of the programmable machine, and how does this connect to today's AI?
The Baroque approach consisted of collecting ordered facts within specific domains of knowledge, whereas the 19th-century concept introduced programmability and the separation of operations from the object of those operations. Modern AI continues this direction, yet it creates the risk of confusing external signs of competence with actual possession of knowledge.
How did the transition from the calculator to the Analytical Engine change our understanding of information processing?
This transition enabled the separation of the device's material construction from the function it performs through variable rules provided from the outside. As a result, information processing ceased to be identified exclusively with arithmetic, and numbers became carriers of codes representing any symbols and relations.
How do the mechanism of punched cards and the views of Ada Lovelace relate to the contemporary debate on AI autonomy?
The mechanism of punched cards shows that simple instructions can create complex patterns at a higher level of organization, which relates to the modern problem of the unpredictability of AI system results. On one hand, Ada Lovelace claimed that the machine cannot independently invent new things; on the other, she recognized that its operation could lead humans to previously unknown results. Consequently, Lovelace represents both sides of the debate: she is the author of the argument against the creative autonomy of machines, while simultaneously recognizing the potential of systems to generate surprising outcomes.
How can machines process knowledge about the world without having direct contact with it?
Machines can process knowledge about the world by operating on symbolic systems and representations created by humans, such as models, data, or vectors. Instead of direct contact with reality, these devices utilize recorded dependencies and parameter structures, which allow for useful operations to be performed without the need for physical experience of the world.
How does a computer program change the relationship between a symbol (text) and an action in the material world?
A computer program makes the record a mechanism of action, and the machine becomes a text that can be edited. Thanks to programmability, an instruction ceases to be merely a description and begins to act causally on the machine, directly controlling matter.
How did the mechanization of industrial labor affect the way texts are created and how creativity is thought about?
Mechanization led to the emergence of so-called Template Culture, in which texts and intellectual work began to be treated as processes capable of standardization, replication, and production according to a pattern. Creativity ceased to be perceived exclusively as the result of individual talent and began to be understood as an activity based on repeatable structures, procedures, and tools.
How is the historical division of labor in industry related to the functioning of modern artificial intelligence?
The historical division of labor enabled the breaking down of complex cognitive competencies into simple operations, which became a necessary condition for subsequent automation. Modern artificial intelligence functions as a technical consolidation of the results of this collective work by many people, whose knowledge and practices have been objectified in training data.
How do forms and established writing patterns influence how we think and communicate in society?
Templates and rhetorical genres organize social communication by imposing specific forms, tones, and scopes of information, which reduces cognitive costs for both the sender and the receiver. By ensuring the repeatability and predictability of utterances, they allow for the quick identification of a text's purpose and make the functioning of institutions independent of the personal characteristics of the individual performing the task.
Does AI introduce a new form of creative automation, or does it merely accelerate the standardization processes that have long been present in culture?
AI does not create the phenomenon of co-authorship from scratch; rather, it densifies, accelerates, and makes visible a process that already existed in the form of genre conventions, recording technologies, or rhetorical traditions. The formalization of the creative process does not necessarily eliminate creativity, but merely changes its point of application, shifting the author's role toward the selection and modification of provided structures.
How does AI affect the structure of intellectual work, and does it lead to the mass replacement of humans?
AI affects the structure of work through so-called cognitive Taylorism, which consists of extracting the dispersed knowledge of employees and transforming it into normative patterns and standards that govern their work. Instead of mass human replacement, a transformation of professions is more likely, as empirical data do not show labor displacement on the scale of catastrophic forecasts.
What is the greatest risk of automating cognitive processes, and what does it mean to be an author in this context?
The greatest risk of automation is a situation where patterns become so invisible and ubiquitous that we begin to mistake them for the natural structure of thinking, which may define the boundaries of what we consider a reasonable possibility. In this context, being an author is not about creating without help, but about the ability to recognize a template and consciously enter into a relationship with it by accepting, transforming, or rejecting it.
