A carregar...
Publicação
Inteligência artificial e consciência fenoménica : quão perto estamos de máquinas conscientes?
| Nome: | Descrição: | Tamanho: | Formato: | |
|---|---|---|---|---|
| 1.78 MB | Adobe PDF |
Autores
Orientador(es)
Resumo(s)
A consciência é um fenómeno mal compreendido de sistemas biológicos com uma certa complexidade, e há quem sustente a possibilidade de poder vir a ser uma propriedade de determinados organismos artificiais. Alguns autores acreditam que pode emergir do substracto físico, ser uma ilusão ou um mero epifenómeno pelo que, teoricamente, nada obsta a que possa ser instanciada, uma vez compreendidos os mecanismos que a fazem surgir. Para outros, entre os quais me incluo, a instanciação da consciência em organismos artificiais não é possível, uns porque a consideram irredutível ao físico, outros porque a situam em planos transcendentais.
Neste trabalho, para além de considerações gerais destinadas a situar o assunto, procuro abordar particularmente o conceito de consciência fenoménica, o chamado “problema duro”, o problema de saber como é que certas actividades neuronais aparecem internamente como experiência subjectiva, como qualia, à luz de diferentes teorias oriundas de vários campos da ciência. Para além das referências às principais teorias metafísicas, discuto com algum pormenor as mais relevantes teorias específicas da consciência, analiso modelos e implementações propostas pela Inteligência Artificial (IA) e pela Consciência Artificial (CA)1, discuto até que ponto se avançou, ou não, na simulação e na instanciação da consciência em organismos artificiais, e quais as principais objecções à sua instanciação e caracterização.
Por fim são extraídas algumas conclusões que tentam responder à questão suscitada no título, partindo da ideia de que a consciência fenoménica não é processamento de informação, de uma intuição a priori2 de que não é uma propriedade emergente do substracto físico, e que talvez só seja possível em determinados organismos biológicos.
Consciousness is a poorly understood phenomenon of both more and less complex biological systems, and there are those who defend the possibility that it may also become a property of certain artificial organisms. Some authors believe that consciousness can arise from the physical substrate, be an illusion or a mere epiphenomenon, so theoretically nothing prevents it from being implemented once the mechanisms that make it emerge are understood. For others, myself included, the implementation of consciousness in artificial organisms is not possible, either because it is considered to be irreducible to the physical, or because it is situated in transcendental order. In this work, in addition to general considerations aimed at placing the subject, I particularly seek to address the concept of phenomenic consciousness, the so-called "hard problem," the problem of how, in the light of different theories from various fields of science, certain neural activities appear internally as subjective experience, or qualia. In addition to references to the main metaphysical theories, I discuss in some detail the most relevant specific theories of consciousness, analyze models and implementations proposed by Artificial Intelligence and Artificial Consciousness, and discuss to what extent progress has been made in the simulation and instantiation of consciousness in artificial organisms, and what are the main objections to its instantiation and characterization. Finally, some conclusions are drawn in an attempt to answer the question raised in the title, starting from the idea that phenomenic consciousness is not a processing of information, and from an a priori intuition that it is not an emergent property of the physical substrate, and that perhaps it is only possible in certain biological organisms.
Consciousness is a poorly understood phenomenon of both more and less complex biological systems, and there are those who defend the possibility that it may also become a property of certain artificial organisms. Some authors believe that consciousness can arise from the physical substrate, be an illusion or a mere epiphenomenon, so theoretically nothing prevents it from being implemented once the mechanisms that make it emerge are understood. For others, myself included, the implementation of consciousness in artificial organisms is not possible, either because it is considered to be irreducible to the physical, or because it is situated in transcendental order. In this work, in addition to general considerations aimed at placing the subject, I particularly seek to address the concept of phenomenic consciousness, the so-called "hard problem," the problem of how, in the light of different theories from various fields of science, certain neural activities appear internally as subjective experience, or qualia. In addition to references to the main metaphysical theories, I discuss in some detail the most relevant specific theories of consciousness, analyze models and implementations proposed by Artificial Intelligence and Artificial Consciousness, and discuss to what extent progress has been made in the simulation and instantiation of consciousness in artificial organisms, and what are the main objections to its instantiation and characterization. Finally, some conclusions are drawn in an attempt to answer the question raised in the title, starting from the idea that phenomenic consciousness is not a processing of information, and from an a priori intuition that it is not an emergent property of the physical substrate, and that perhaps it is only possible in certain biological organisms.
Descrição
Palavras-chave
Inteligência artificial Consciência Cognição Qualia Experiência Intuição Ciências cognitivas Teses de mestrado - 2018