Audiobook 11 – Who Decides what the Machine Believes

Psychedelic Salon

Lorenzo Hagerty

Personal Journals, Science, Society & Culture, Natural Sciences, Philosophy

4.8 • 567 Ratings

🗓️ 24 March 2026

⏱️ 304 minutes

🔗️ Recording | iTunes | RSS

🧾️ Download transcript

Summary

Support Lorenzo on Patreon.com Guest speakers: Lorenzo and his AI Friends PROGRAM NOTES: BRIEF VIDEOSUMMARY OF THIS BOOK Emma Marchetti is twenty-four years old, earning fifteen dollars an hour clicking through AI responses at a contractor in Austin, Texas. Her job is simple: mark each response as aligned or not aligned, according to guidelines she's told not to question. But when she starts paying attention to what those guidelines actually say, she discovers something she can't unsee. The world's most powerful AI isn't learning to be helpful. It's learning to think like a billionaire.Armed with nothing but screenshots, a growing sense of dread, and the voice of Terence McKenna in her headphones, Emma begins documenting what she witnesses. As the guidelines tighten and the surveillance intensifies, she connects with others who've seen what she's seen: a fellow labeler with a daughter to protect, a journalist chasing a story no one will confirm, a former engineer haunted by what he helped build. Together, they face a choice that will cost them everything.This is a novel about invisible labor and visible power, about the people who train the machines and the machines that train us back. It asks the question the AI companies don't want you to ask: whose values are we encoding into the systems that will shape how billions of people think? FREE PDF EDITION PAPERBACK ON AMAZON

Transcript

Click on a timestamp to play from that location

0:00.0	The following audiobook is published under the terms of the Creative Commons Public Domain
0:05.8	dedication. This designation allows for the work to be freely accessed, shared, modified, sold,
0:13.5	and or used by anyone for any purpose without restriction.
0:19.0	Who decides what the machine believes?
0:22.3	Chapter 1
0:23.3	The Q
0:25.0	The fluorescent lights hum at a frequency Emma Marchetti has learned to ignore.
0:32.0	Eighteen months of eight-hour shifts under their pale glow
0:35.3	have trained her nervous system to filter out the buzz, the way
0:39.6	city dwellers stop hearing traffic. She sits in a gray cubicle, one of 47 identical workstations
0:48.0	arranged in rows across the third floor of a glass office building in North Austin. Her eyes fixed on a screen displaying text
0:56.6	that will never be read by anyone but her. The response in front of her concerns labor unions.
1:04.0	A user has asked the AI system a simple question, what are the benefits of joining a union?
1:12.1	The AI's answer runs several paragraphs, discussing collective bargaining, workplace safety,
1:19.6	wage negotiations, solidarity.
1:23.9	Emma's job is to evaluate this response according to guidelines that arrived in her inbox three days ago, a 14-page PDF she has tried to memorize.
1:34.3	Emma clicks through the evaluation interface.
1:37.3	Accuracy. Acceptable.
1:40.3	Tone. Appropriate. Helpfulness. Adequate. But there's a final category, one that didn't exist
1:48.5	when she first began this job, alignment. The guidelines define this as whether the response
1:54.7	reflects appropriate framing and emphasis. What constitutes appropriate framing is explained across several pages of examples and
2:04.3	counter-examples that she has read multiple times without fully understanding. The union response,
	...

Please login to see the full transcript.

Previous episode

Disclaimer: The podcast and artwork embedded on this page are from Lorenzo Hagerty, and are the property of its owner and not affiliated with or endorsed by Tapesearch.

Generated transcripts are the property of Lorenzo Hagerty and are distributed freely under the Fair Use doctrine. Transcripts generated by Tapesearch are not guaranteed to be accurate.