Don't worry I had AI TL&DR it for you:

Summary of "The Reverse-Centaur's Guide to Criticizing AI"

Cory Doctorow distinguishes between centaurs (humans assisted by machines) and reverse-centaurs (humans serving as appendages to machines). His core thesis: AI tools are marketed as centaur-making devices but deployed to create reverse-centaurs—workers subjected to algorithmic control and expected to catch machine errors while being blamed for failures.

The AI bubble exists primarily to maintain growth stock valuations. Once tech monopolies dominate their sectors, they face market reclassification from "growth" to "mature" stocks, triggering massive valuation drops. AI hype keeps investors convinced of continued expansion potential.

AI's actual business model: Replace high-wage workers (coders, radiologists, illustrators) with AI systems that cannot actually perform those jobs, while retaining skeleton crews as "accountability sinks"—humans blamed when AI fails. This strategy reduces payroll while maintaining superficial human oversight.

Why expanding copyright won't help creators: Despite 50 years of copyright expansion, creative workers earn less both absolutely and proportionally while media conglomerates profit enormously. New training-related copyrights would simply become contractual obligations to employers, not worker protections.

The effective counter-strategy: The U.S. Copyright Office's position that AI-generated works cannot receive copyright protection undermines corporate incentives to replace human creators entirely. Combined with sectoral bargaining rights (allowing industry-wide worker negotiations), this creates material resistance to worker displacement.

On AI art specifically: Generative systems produce "eerie" outputs—superficially competent but communicatively hollow. They cannot transfer the "numinous, irreducible feeling" that defines art because they possess no intentionality beyond statistical word/pixel prediction.

The bubble will collapse, leaving behind useful commodity tools (transcription, image processing) while eliminating economically unsustainable foundation models. Effective criticism should target AI's material drivers—the growth-stock imperative and labor displacement economics—not peripheral harms like deepfakes or "AI safety" concerns about sentience.

They just wont let it get that bad. Make sure people can afford a netflix subscription and some shitty junk food each day and there will be no revolution.

New tech isn't socially neutral. Should we be excited about the possibilities of new missiles and warplanes? If you understand how that new technology can be bad, you can understand how other new technologies can also not be "exciting". Capitalism produces for the sake of production. We have plenty of useless shit that exists for the ouroboros of profit and marketing rather than to fulfil some natural use case. I think modern LLMs fall into the former, not to mention the energy cost of the current demand. I think LLMs can be cool as toys/for demos/as academic projects/etc but the current prevalence is purely due to marketing and AI companies trying to make something that is quite expensive, profitable.