MindMap Gallery week 2 media texts
This mind map elaborates on communication theory concepts such as lecture semiotics for beginners, semiotics, and medium. By explaining these concepts, the map reveals the frameworks and methodologies of semiotics in interpreting and understanding cultural phenomena.
Edited at 2021-10-11 11:18:14Week 2 Media Texts
Lecture Semiotics for Beginner (Intro) Daniel Chandler
-表示 -内涵 -神话 -范例 -句法
-诠释学家 -动作语码 -消音代码(隐含含义) -符号代码 -文化代码
-Social Semiotc社会符号学 - Three Mentafunction心理功能
-多模态的 -Discourse话语
Orchestrating:编排 Contributory:贡献 Embodied:提现
Definition: A science which studies the role of signs as part of social life.
Lecture Semiotics for Beginner (Intro) Daniel Chandler
Semiotics
Definition: A science which studies the role of signs as part of social life.
It would form part of social psychology, and hence of general psychology.
Investigate the nature of signs and the laws governing them.
Structuralism
Structuralism is an analytical method which has been employed by many semioticians
Shape the epistemic boundaries governing how AI systems operate
Modern semiotic theory is also sometimes allied with a Marxist approach which stresses the role of ideology
Semiotics began to become a major approach to cultural studies in the late 1960s
Barthes declared that 'semiology aims to take in any system of signs, whatever their substance and limit, which form the content of ritual, convention or public entertainment
Medium
Human experience is inherently multisensory, and every representation of experience is subject to the constraints and affordances of the medium involved.
More frequently and fluently a medium is used, the more 'transparent' or 'invisible' to its users it tends to become.
The choise of medium leads to influences that user may not be conscius
Umberto Eco:" Semiotics is concerned with everythin that can be taken as a sign"
Seminar Excavating AI
Introduction
The automated interpretation of images is a social and political project, rather than a purely technical one.
What work do images do in AI systems?
What is misrecognized or even completely invisiblene.
The method for introducing images into computer systems & How taxonomies order the foundational concepts that will become intelligible to a computer system.
How do humans tell computers which words will relate to a given image?
What is at stake in the way AI systems use these labels to classify humans, including by race, gender, emotions, ability, sexuality, and personality?
That computer vision is meant to serve in our society—the judgments, choices, and consequences of providing computers with these capacities.
Training AI
Building AI systems requires vast amount of data
Training sets are the foundation on machine-learning systems
Shape the epistemic boundaries governing how AI systems operate
The image used in computer-vision systems composed of shaky assumptions
Images are remarkable, has multiple potential meanings, irresolvable questions, and contradictions. Therefore, entire philosophy, history and media theory are decicated to teasing out the nuances of the relationship between images and meanings.
Image do not decribe themselves
It means different things depending on who looks and how it located.
Image are open to interpretation and reinterpretation
Anatomy of a Training Set
Training sets (collection of images) has been labeled in various ways and categories
Three layers:
Overall Taxonomy分类系统( Aggregate rules)
Individual Classes (Fruit)
Each Individually Labeled Image (Apple)
The Canonical Training Set: ImageNet
Idea: Map out the entire world of objects
The underlying structure of ImageNet is based on the semantic structure of WordNet, a database of word classifications
Taxonomy
Nested Structure of Cognitive Synonyms认知同义词的嵌套结构
“synset” represents a distinct concept, with synonyms grouped together (Apple & Orange) → “Synset" organized into a Hierarchy
Eg.“chair” is nested as artifact > furnishing > furniture > seat > chair
Categories
Divide an infinitely complex universe into separate phenomena
To impose order onto an undifferentiated mass
No gradient in the logic of ImageNet. Everything is flattened out and pinned to a label.
Assemble vast photo from people without consent or participation
ImageNet Roulette: An Experiment in Classification
Train an AI model exclusively on one category
Shed light on consequence when technical systems are trained using problematice data
Assemble vast photo from people without consent or participation
Labeled Images
Divide an infinitely complex universe into separate phenomena
To impose order onto an undifferentiated mass
No gradient in the logic of ImageNet. Everything is flattened out and pinned to a label.
Assemble vast photo from people without consent or participation
ImageNet Roulette: An Experiment in Classification
Images are laden with potential meanings, irresolvable questions, and contradictions
Shed light on consequence when technical systems are trained using problematice data
Assemble vast photo from people without consent or participation
relationship between picture and concept recall physiogonomy人相学,the pseudoscientific伪科学的 assumption that something about a person's character can be gleaned by observing their features.
UTK: Making Race and Gender from Your Face
Objects perserve mathematically their forms
Created composite images of criminals, studied the feet of prostitutes, measured skulls, and compiled meticulous archives of labeled images and measurements
Capture and pathologize what was seen as deviant or criminal behavior, and make such behavior observable in the world
At the level of the image label is the assumption that someone’s gender identity can be ascertained
Eg. Classificatory schema: racial classification
Improve matters by producing “more diverse” AI training sets presents its own complications.通过生产“更多样化”的人工智能训练集来改善问题本身就是其复杂性。
IBM’S Diversity in Faces
Created as responese that can't recognize face of darker skin
“Diversity in Faces” (DiF) dataset
“a computationally practical basis for ensuring fairness and accuracy in face recognition,
The dataset itself continued the practice of collecting hundreds of thousands of images
including facial symmetry and skull shapes to build a complete picture of the face
too the technical process of categorizing and classifying people is shown to be a political act
Using mathematical approch to quantify "diversity" and "evenness"均匀性
Expanding dataset can only imporve the accuracy but cannot effect the final classification.
Achieving parity amongst different categories is not the same as achieving diversity or fairness,
Epistemics of Training Sets
Undergriding Assumption
Underlying Theoretical Paradigm assumes Concept (Allow transcendental grounding and internal consistency)
Assumes a fixed correspondences between images and concepts, appearances and essences.
Visual essence视觉本质 is discernible识别 by using statistical methods to look for formal patterns across a collection of labeled images.
The datasets of AI is built with unstable epistemological and metaphysical assumption认识论和玄学的假设
Datasets aren’t simply raw materials to feed algorithms算法, but are political interventions. (Bias)
Collecting Image, Categorzing, Lable is decided from sicial and political work.
Missing Person
Lots facial datasets that without consent was removed as they are obvious privacy and ethical violations.
Useless: Countless download and Production of AI
Propagating into hiring, education and healthcare.
Part of security of airport
Consequence:Researchers cannot see the assumption, lables and classificatory that in new systems
Developing frameworks within which future researchers can access these data sets in ways that don’t perpetuate harm is a topic for further work
Conclusion: Who decides
Physiognomists: Relationship between Image of person and the Character was inscribed.刻
Magritte: Images have unstable relationship to what they present.
Image has much broader politics of representation and self-representation.
Meaning of Image is basis architecture and contents of training sets for AI, which promote or discriminate, approve or reject, render visible or invisible, judge of enforce.
Future: 1. We need to examine them—because they are already used to examine us 2. Having wider public discussion about the consequences rather than keeping it in academic corridors.
Underexamined role: the power to shape the world in their own images.