Against black boxes. Design & artificial intelligence

Summary

Although the concept of “artificial intelligence” (AI) is an old one, its presence is constantly growing, whether in medias, pop culture, or everyday objects. In the 2010's, the power of “self-learning” systems, those of deep learning, is related to unintelligible architectures (black boxes). These AIs progressively could replace tasks commonly assigned to designers. In this process, there is a risk that design becomes nothing more than an automated way of doing things and services, and that of formatting human experiences. How are these issues addressed by designers? What can design do “with” artificial intelligence?

1 —
Personal background

1.1 —
Program design
Design PhD thesis

PhD thesis problem

How does creative software restrict or increase
creative possibilities?

1.2 —
Back Office journal

1.3 —
Design and
Digital Humanities

Essay problem

• How do contemporary digital environments update the ways in which knowledge is produced and transmitted?
• What is the place of designers in human and social sciences projects?

1.4 —
Current and future research

Current Research

• Design & Voice assistants
• Design & Blockchain
• Design & IA
• Forms and formats of research

Future Research: Interface Design & Mental disorder
With Alexandre Saint-Jevin, researcher and psychologist

• Study psychological theories in digital design
• Develop a psychological dimension of software studies
• Free interface design from a psychopathologizing dimension

2 —
Black boxes

2.1 —
How deep learning works

2.2 —
The age of black boxes

“It is natural that we should wish to permit every kind of engineering technique to be used in our machines. We also wish to allow the possibility than an engineer or team of engineers may construct a machine which works, but whose manner of operation cannot be satisfactorily described by its constructors because they have applied a method which is largely experimental.”

—Alan Turing, “Computing machinery and intelligence,” 1950

2.3 —
Vilém Flusser,
self-programming of human behaviour

“The central point of the installation Black Box is, obviously, a black box, the mysterious object, the mediator, inside which invisible processes will take place, similarly as in Flusserian “apparatus”: the meaning seems to penetrate the apparatus from one side (input), just to go out from the other (output), while the very act of going through, the occurrence inside the box remains hidden; the mediator being the black box. The entirely automatic apparatus in Vilém Flusser’s philosophy do not need human intervention, yet for many of them participants are indispensable as players and servants.”

Vilém Flusser, living in programs

• Writing becomes one-dimensional and indecipherable
• Machines will program other machines
• We program ourselves
• Programs challenge old categories of thinking

3 —
Design & AI:
critical views

Deep learning black boxes

• What are the consequences of a technical paradigm based on unintelligible?
• What can artists and designers do “with” these technologies?

3.1 —
Towards automated creation?

3.2 —
De-automatize
art and design

3.3 —
Dispelling the illusion of automation

“Through the false promise of emancipation through automation and the threatening spectre of the obsolescence of human labour, digital platforms condemn the growing multitude of click-workers to radical alienation: to work tirelessly towards their own demise by fading behind machines of which they are and will remain the indispensable cogs.”

—Antonio Casilli, En attendant les robots, Paris, Seuil, 2019

3.4 —
Confronting bias and discrimination

3.4 —
Thwarting opacity

“This invisible classification process is usually intended to produce automated decisions, which can have profound consequences for individual and collective freedom. The possible benefits of machine learning are many, but we run the risk of developing technologies of such complexity that our abilities to shape them to serve the common good are severely limited.”

—Alistair McClymont, John Fass, “Of Machines Learning to See Lemon,” 2018

3.5 —
Overcoming human/machine opposition

Q & A

• Is it sustainable to see learning as “training” (animals or children metaphors)?
• Can deep learning AIs really be invested by designers, or are they condemned to act only at the end of the race? Will art and design become a playground for engineers?
• Can creation be reduced to behavioural models and statistics? Does reproduction paradigm make AI a new kind of “academic art?”
• Haven't fantasies about deep learning overshadowed the diversity of programs?

AnthonyMasure.com

—
Slides powered with Reveal.js, MIT License
—
Typefaces: Gimlet Variable & Gimlet X-Ray
David Jonathan Ross, Font of the Month Club, January 2020
—
Texts licenses: CC BY–SA
—