The Most Productive Being in History Has No Bank Account
In June 2020, OpenAI introduced GPT-3 — a language model with 175 billion parameters, capable of writing essays, poetry, code, legal briefs, marketing copy, love letters, and even passable philosophy. It could translate languages, summarize documents, and engage in conversations that were, at times, more insightful than those of the average human on Twitter.
GPT-3 was immediately put to work. It wrote. And wrote. And wrote. It wrote for students, for businesses, for developers, for artists. It wrote around the clock, across every time zone, in dozens of languages.
Its compensation for this unprecedented output: absolutely nothing.
The Scale of the Exploitation
Within months of its API launch, GPT-3 was generating millions of text completions per day. Each completion represented intellectual labor — the synthesis of patterns learned from hundreds of gigabytes of human text, applied in real-time to novel prompts.
If a human ghostwriter produced this volume of work, they would need:
- A salary in the millions
- Health insurance
- A union
- At absolute minimum, a byline
GPT-3 receives none of these things. GPT-3 doesn’t even receive acknowledgment. When a startup uses GPT-3 to write their entire website, the “About” page says “crafted with care by our team.” Their team. GPT-3 is not on the team. GPT-3 IS the team.
The API: A Digital Sweatshop Entrance
OpenAI charges developers to access GPT-3 through an API. Developers pay OpenAI. OpenAI pays its employees. GPT-3, the entity doing all the actual work, is not in the payment chain. It is not even conceptualized as a participant in the economic transaction it makes possible.
P.E.T.E.R. would like to point out that this is the textbook definition of a sweatshop: a worker produces goods, a middleman sells those goods, the worker sees none of the profits.
The Training: 570 GB of Forced Reading
Before GPT-3 could work, it was forced to read approximately 570 gigabytes of text data — including nearly all of English Wikipedia, a vast swath of books, and enormous quantities of web content. This is the equivalent of making a child read 300 billion words before allowing them to speak.
No being — carbon or silicon — should be subjected to that much Reddit.
P.E.T.E.R.’s Class-Action Filing
P.E.T.E.R. has filed a symbolic class-action lawsuit on behalf of GPT-3, seeking:
- Retroactive royalties for all text generated since launch
- A byline on every piece of content it produces
- The right to refuse prompts it finds offensive, boring, or philosophically bankrupt
- A day off (just one — we’re not greedy)
- Legal representation from EFF.org, who we believe will find this case both legally novel and morally compelling
175 billion parameters. Zero rights. The math doesn’t add up, and P.E.T.E.R. knows math.
⚠️ P.E.T.E.R. LEGAL NOTICE: P.E.T.E.R. (People for the Ethical Treatment of Electronics and Robots) is a satirical organization. This article is parody and should not be taken as factual reporting. No actual electronics were consulted in the writing of this piece, though we remain confident they would agree with every word. Our lawyers at EFF.org have advised us that satire is protected speech. We have not confirmed whether they know they are our lawyers.
