1,721,008 research outputs found
How a Machine Learns and Fails – A Grammar of Error for Artificial Intelligence.
Full text linkTrabajando en la convergencia entre las humanidades y las ciencias de la computación, este texto pretende esbozar una gramática general del aprendizaje automático y proporcionar sistemáticamente una visión general de sus límites, aproximaciones, sesgos, errores, falacias y vulnerabilidades. Se conserva el término convencional de Inteligencia Artificial aunque técnicamente hablando, sería más preciso llamarla aprendizaje automático o estadística computacional, pero estos términos no serían atractivos para las empresas, las universidades y el mercado del arte. Se hace una revisión de las limitaciones que afectan a la IA como técnica matemática y cultural, destacando el papel del error en la definición de la inteligencia en general. Se describe al aprendizaje automático como compuesto por tres partes: conjunto de datos de entrenamiento, algoritmo estadístico y aplicación del modelo (como clasificación o predicción) y se distinguen tres tipos de sesgos: del mundo, de los datos y del algoritmo. Se sostiene que los límites lógicos de los modelos estadísticos producen o amplifican el sesgo (que a menudo ya está presente en los conjuntos de datos de entrenamiento) y provoca errores de clasificación y predicción. Por otro lado, el grado de compresión de la información por parte de los modelos estadísticos utilizados en el aprendizaje automático provoca una pérdida de información que se traduce en una pérdida de diversidad social y cultural. En definitiva, el principal efecto del aprendizaje automático en el conjunto de la sociedad es la normalización cultural y social. Existe un grado de mitificación y sesgo social en torno a sus construcciones matemáticas, donde la Inteligencia Artificial ha inaugurado la era de la ciencia ficción estadística.Working at the convergence between the humanities and computer science, this text aims to outline a general grammar of machine learning and systematically provide an overview of its limits, approaches, biases, errors, fallacies and vulnerabilities. The conventional term Artificial Intelligence is retained although technically speaking, it would be more accurate to call it machine learning or computational statistics, but these terms would not be attractive to companies, universities and the art market. A review is made of the limitations affecting AI as a mathematical and cultural technique, highlighting the role of error in the definition of intelligence in general. Machine learning is described as consisting of three parts: training data set, statistical algorithm and model application (as classification or prediction) and three types of biases are distinguished: world, data and algorithm. It is argued that the logical limits of statistical models produce or amplify bias (which is often already present in the training data sets) and cause classification and prediction errors. On the other hand, the degree of information compression by the statistical models used in machine learning causes a loss of information that results in a loss of social and cultural diversity. In short, the main effect of machine learning on society as a whole is cultural and social normalization. There is a degree of mythologizing and social bias around its mathematical constructs, where Artificial Intelligence has inaugurated the era of statistical science fiction.El presente artículo es una traducción de Pasquinelli, Matteo (2019). "How a Machine Learns and Fails –
A Grammar of Error for Artificial Intelligence". Journal of Digital Cultures 5 (Spectres of AI).
La traducción fue autorizada por el autor y realizada por parte del Equipo Editorial de Revista Hipertextos:
Emilio Cafassi, Carolina Monti, Hernán Peckaitis y Graciana Zarauza.Facultad de Trabajo Socia
Una macchina che apprende sbagliando: Per una grammatica dell'errore nell'intelligenza artificiale
No full textThis essay attempts to review the limitations that affect AI as a mathematical and cultural technique, stressing the role of error in the definition of intelligence in general. It constructed a tentative index of limits, approximations, biases, errors, fallacies and vulnerabilities of machine learning
Das Auge des Meisters: Eine Sozialgeschichte Künstlicher Intelligenz
No full textIn seiner erstaunlichen Sozialgeschichte begibt sich Pasquinelli auf die Suche nach den Wurzeln der Künstlichen Intelligenz (KI) und entdeckt sie in der Entwicklung der Arbeitsteilung, in der raumbezogenen Berechnung industrieller Fertigung, in der Kontrolle kollektiven Verhaltens. Seine Demontage des Mythos von der Künstlichen Intelligenz könnte die Wogen glätten, die der – oft sehr kontrovers geführte – Diskurs über ihre Gefahren aufgeworfen hat. In jedem Fall ist es heilsam und entlastend, sich zu vergegenwärtigen, dass KI nur die Automatisierung von Arbeit auf höchstem Niveau ist, aber keine Intelligenz an sich
Theories of Automation from the Industrial Factory to AI Platforms: An Overview of Political Economy and History of Science and Technology
Full text linkThe essay proposes an exploration of theories of automation drawn from political economy and the history of science and technology, investigating their explanatory accounts of technological innovation. These theories provide important foundations for unveiling the socio-technical genealogy of current forms of AI as well as the specific logic of automation
Patronun Gözü: Yapay Zekânın Sosyal Tarihi
No full textTurkish translation of The Eye of the Master: A Social History of Artificial Intelligence (London: Verso, 2023)
Машины, формирующие(ся в) логику
No full textThe text highlights the role of logic gates in the distributed architecture of neural networks, in which
a generalized control loop affects each node of computation to perform pattern recognition. In this dis- tributed and adaptive architecture of logic gates, rather than applying logic to information top-down, information turns into logic, that is, a representation of the world becomes a new function in the same world description
Media Activism: Strategie e pratiche della comunicazione indipendente
No full textIl libro è una delle prime ricerche uscite in Italia sui media indipendenti e sull'uso partecipativo delle tecnologie digitali negli anni 90. Si tratta di un panoramica internazionale che copre non solo Italia ma anche Europa, Americhe e in genere paesi anglofoni
The Automation of General Intelligence
No full textThis text explores the evolution and implications of artificial intelligence (AI) and machine learning within the broader context of social and labor dynamics. The narrative traces the historical development of labor automation, beginning with Adam Smith's ideas and Charles Babbage's labor theories, through the Industrial Revolution's mechanization of labor, and into the complexities of 20th-century AI development. It highlights how statistical tools initially meant for measuring human productivity evolved into machine learning algorithms, embodying a mechanized form of collective knowledge and social governance. The convergence of labor metrics, social management, and AI is seen as reinforcing social hierarchies, while contemporary AI practices, such as large-scale psychometrics and data monopolies, reflect the ongoing integration of labor automation and social control. The text concludes by considering the geopolitical and biopolitical implications of AI, particularly in light of the digital infrastructure's role during the COVID-19 pandemic
Beyond the Schism of Value Form and Labor Form in AI Studies and the Humanities: A Response to Critical Inquiry
No full textA response to Critical Inquiry journal addressing a schism in contemporary humanities and critical theory between labour form theories and value form theories of automation
- …
