Globalization made the world flat, AI drills holes into the infinitesimally small

The function of deep learning

Digital oil?

The function of deep learning

Digital oil?

The function of deep learning

Digital oil?

The function of deep learning

Digital oil?

Data as Oil

Own your data
Legislation for the flow of data
Ethically source data
Don't reinforce biases

The function of deep learning

Digital oil?

The function of deep learning

Digital oil?

WhatWhy is data?

The function of deep learning

Digital oil?

WhatWhy is data?

collected

The function of deep learning

Digital oil?

The function of deep learning

Digital oil?

"[W]e created a new web scrape which emphasizes document quality." "To do this we only scraped web pages which have been curated/filtered by humans." "Manually filtering a full web scrape would be exceptionally expensive so as a starting point, we scraped all outbound links from Reddit, a social media platform, which received at least 3 karma." "The resulting dataset, WebText, contains the text subset of these 45 million links."

-Alex Radford et. co, Language Models are Unsupervised Multitask Learners (2019)

The function of deep learning

Digital oil?

Data is work.

The function of deep learning

Digital oil?

The function of deep learning

Digital oil?

Data does not burn.

The function of deep learning

Digital oil?

This is why stealing from one artist, the network stole from every other artist.

It stole from every artist that served as inspiration, from every artist they'll inspire, from every artist that approaches the same subject or mood and from every artist they have nothing in common with, because this is how the network functions.

There is more still.

The network also stole from every art historian developing periodizations, structures and hierarchies between works, from every critic that developed the language we use to describe images, from every teacher that diffused that language, from every Wikipedia contributor, it stole from everyone who ever posted a captioned image on social media.

It stole quanta of time, so small they can't be registered, but it did so from millions, maybe billions of people, over decades.

It stole quanta of time that are themselves the result of centuries of knowledge production and social practices, many thought to be constrained to the public sphere, offered for the general wellbeing.

Each and every one of us has a claim over something like ChatGPT.

Some might have a larger one than others, but only in the sense that a nanoscopic claim is larger than a picoscopic one.

The claims are so thin and interdependent and unpredicted that they can't be resolved.

This is why, adequately addressing AI cannot be done in a framework that sees ownership as absolute instead of transient, a framework that rejects the social and the public, in favor of the personal and private.

Ending remarks

AI will destroy mankind or allow it to destroy itself
AI will eliminate most jobs - good
AI will eliminate most jobs - bad
Nothing transformative will happen

Ending remarks

Never let a good crisis go to waste.

But we do find ourselves in a moment of crisis that requires more profound technological change at the level of production than what has been happening for the past 40 years.

Subtly, AI will be part of that change, as a way to harness the general intellect of humanity.

That is not why AI is important right now.

It is important because there is no way we can approach it meaningfully as individuals. Neither is war, rampant inequality, a pandemic or climate change.

From this point of view, AI is a blessing in disguise.

It illustrates the network of dependencies that flow through each and every one of us, it makes quantifiable the ties that bind us.

We just have to sharpen our vision enough to see them.

And we have to do it now, when it feels like we are living through an apocalypse, when the veil between the worlds can be slit.

Not before long, the moment of Revelation will pass and we'll find ourselves living in a new world shaped by its fires.

References and further reading

Glazer, Nona. Servants to Capital: Unpaid Domestic Labor and Paid Work

Mandel, Ernest. Long Waves of Capitalist Development

Marx, Karl. Capital Vol.1, 2, 3

Rubin, Isaak Ilich. Essays on Marx's Theory of Value

Berlyand, Leonid & Jabin Pierre-Emmanuel. Mathematics of Deep Learning

Radford, Alec. Learning Transferable Visual Models From Natural Language Supervision

Ramesh, Aditya et all. Hierarchical Text-Conditional Image Generation with CLIP Latents

Weidman, Seth. Data Learning From Scratch

Fine, Ben.
Microeconomics: A Critical Companion

Smith, Jason.
Smart Machines and Service Work

Wasielewski, Amand.
Computational Formalism

Aguilera, Rodrigo.
The Glass Half-Empty

Hadfield, Gillian.
Rules For a Flat World

Hickel, Jason.
The Divide