• JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
 
  Bookmark and Share
 
 
Doctoral Thesis
DOI
https://doi.org/10.11606/T.47.2019.tde-12122019-163927
Document
Author
Full name
Paula Jacqueline de Oliveira
E-mail
Institute/School/College
Knowledge Area
Date of Defense
Published
São Paulo, 2019
Supervisor
Committee
Malvezzi, Sigmar (President)
Alves, Cecilia Pescatore
Barlach, Lisete
Perez, Erico Renteria
Sant'Anna, Anderson de Souza
Tassara, Eda Terezinha de Oliveira
Title in Portuguese
A dinâmica do conhecimento sob o imperativo da tecnologia e da complexidade: uma análise dos mecanismos de Deep Reinforcement Learning e das suas interfaces com o saber, na contemporaneidade
Keywords in Portuguese
Black box
Ciência
Complexidade
Deep learning
Inteligência artificial
Paradigmas científicos
Tecnologia
Trabalho
Abstract in Portuguese
Esta pesquisa teve o objetivo de analisar, caracterizar e descrever os fenômenos emergentes do imperativo da acumulação tecnológica e da complexidade sobre a dinâmica do conhecimento científico, quando utilizadas técnicas de Inteligência Artificial por mecanismos de Deep Reinforcement Learning (DRL). Tal análise foi elaborada por meio de uma meta-análise da totalidade dos artigos veiculados pela organização Deep Mind e disponibilizados de forma gratuita no seu site. Observouse que tais mecanismos têm sido anunciados como alternativas para a resolução de problemas complexos. Observou-se, no entanto, que são de difícil análise, seja a respeito dos paradigmas sobre os quais são desenvolvidos, ou pelas métricas de performance e processamento. A não interpretabilidade de tais mecanismos mostrouse uma preocupação marcante, na comunidade científica. Parece haver, no entanto, uma tendência de se balizar a sua adoção por princípios de utilidade, ou seja, pela sua expressiva redução das margens de erro características de modelos probabilísticos interpretáveis. Observou-se ainda que tais mecanismos tendem a retroalimentar a tríade Ciência Tecnologia Interesse Comercial, na medida em que o consumo de recursos computacionais, exigindo investimentos extremamente altos, os quais tem sido feitos pelas Big Techs, como são denominadas as 05 (cinco) maiores empresas de tecnologia do mundo Apple, Google, Amazon, Microsofg e Facebook
Title in English
Not informed by the author
Keywords in English
Artificial intelligence
Black box
Complexity
Deep learning
Science
Scientific paradigms
Technology
Work
Abstract in English
This research aimed to analyze, characterize and describe emerging phenomena of the imperative of technological accumulation and complexity on the dynamics of scientific knowledge, when using Artificial Intelligence techniques by mechanisms of Deep Reinforcement Learning (DRL). This analysis was elaborated by means of a meta-analysis of all the articles sent by the organization Deep Mind and made available for free in its site. It has been observed that such mechanisms have been announced as alternatives for solving complex problems. It was observed, however, that they are difficult to analyze, either with respect to the paradigms on which they are developed, or by performance and processing metrics. The non-interpretability of such mechanisms has been a major concern in the scientific community. However, there seems to be a tendency to base their adoption on utility principles, that is, on their expressive reduction of the error margins characteristic of interpretable probabilistic models. It was also observed that such mechanisms tend to feed back the Triad Science-Technology-Business-Interest, insofar as the consumption of computing resources, requiring extremely high investments, which have been made by the Big Techs, ) the world's largest technology companies - Apple, Google, Amazon, Microsofg and Facebook
 
WARNING - Viewing this document is conditioned on your acceptance of the following terms of use:
This document is only for private use for research and teaching activities. Reproduction for commercial use is forbidden. This rights cover the whole data about this document as well as its contents. Any uses or copies of this document in whole or in part must include the author's name.
oliveira_do.pdf (2.13 Mbytes)
Publishing Date
2019-12-12
 
WARNING: Learn what derived works are clicking here.
All rights of the thesis/dissertation are from the authors
CeTI-SC/STI
Digital Library of Theses and Dissertations of USP. Copyright © 2001-2024. All rights reserved.