The Future of Everything Is Lies, I Guess — Engineer Warns AI’s Strength Is Also Its Deception
Overview
A long essay published on aphyr.com and circulating on Lobsters argues that current AI — specifically large language models (LLMs) — are essentially “bullshit machines.” It has been reported that the author is releasing the work in sections over several days and will update PDF/EPUB versions as each part appears. The piece reads less like techno-evangelism and more like an elegy: the writer grew up on Asimov and Clarke, dreamed of helpful, thinking machines, and now finds those dreams oddly tarnished by systems that are astonishingly fluent and disturbingly unconcerned with truth.
Claims and context
The essay frames LLMs as statistical improv engines: they predict likely token sequences and thereby “say yes, and…,” producing outputs that sound plausible while sometimes having no grounding in reality. The author walks readers through the mechanics — training on massive text corpora, a costly one-time fit followed by cheap inference — and points to predictable failure modes like confabulation and shallow “memory” implemented by stuffing conversation history into each prompt. The table of contents signals a wide-ranging worry list: dynamics, culture, information ecology, psychological hazards, safety, work, and new roles for humans. It’s part warning, part thought experiment, and part cultural critique.
Why this matters
Why should anyone care? Because these systems are spreading fast. They’re not just toys for researchers; they’re being woven into search, content creation, customer service, and even political messaging. What do we do when the machines we once idolized are better at sounding convincing than at telling the truth? The essay doesn’t offer tidy solutions — and openly admits it’s a polemic — but it forces a necessary question: can society learn to spot and mitigate convincingly phrased falsehoods before they reshape how we work, think, and trust one another? The tone is equal parts frustration and urgency — a reminder that technological progress without wisdom is a risky experiment.
Comments