Cory Doctorow Critiques the AI Bubble and Automation
- •Cory Doctorow critiques the AI industry’s $1.4 trillion capital expenditure as an unsustainable economic bubble.
- •The book defines a 'reverse centaur' as a worker forced to serve as an accountability sink for machines.
- •Doctorow supports web scraping for training models, arguing that record-keeping remains vital for digital history.
Tech journalist and science fiction author Cory Doctorow examines the current state of artificial intelligence and corporate investment trends in his new book, The Reverse Centaur’s Guide to Life After AI. Doctorow critiques the industry's focus on high-capital 'disruptive' models, arguing that the sector relies on an unsustainable 'growth narrative' to satisfy investors after reaching market saturation. Despite global capital expenditure (CapEx) rising to $1.4 trillion, Doctorow characterizes the current AI wave as the most economically inefficient technology cycle in history, noting that the sector turns over roughly $50 billion annually while requiring total asset replacement every 24 to 30 months.
Doctorow distinguishes between the 'centaur'—a human worker who uses AI as an effective tool to enhance their labor—and the 'reverse centaur,' an individual forced to serve as an 'accountability sink' for automated systems. He contends that many corporate leaders push AI implementation to eliminate human reliance, which he describes as a fantasy of a business without employees. Unlike previous technological shifts where workers often sought to adopt new tools, Doctorow observes that contemporary workplaces increasingly use surveillance to force employee adoption, often at the expense of output quality and worker well-being.
The book further challenges the 'anti-AI' sentiment regarding web scraping. Doctorow argues that training models on public internet data is a socially beneficial activity essential for historical record-keeping and data accessibility. He warns that banning the creation of digital records would ultimately empower large media corporations to control information access, rather than protecting individual creators. Doctorow maintains that while the capital allocation and environmental impact of large foundational models remain problematic, the underlying mathematical techniques of statistical inference in neural networks are not inherently harmful.