The Future of Everything is Lies: Aphyr on How ML Promises to be Profoundly Weird
The Future of Everything is Lies: A Comprehensive Critique of Machine Learning by Aphyr
Kyle Kingsbury (Aphyr), the distributed systems researcher known for the Jepsen test series, has published a sweeping multi-part essay titled "The Future of Everything is Lies" that has rocketed to the top of Hacker News with 567 points and 557 comments — making it one of the most discussed articles of the day.
The Core Argument
Aphyr argues that current ML systems are fundamentally "bullshit machines" — sophisticated improvisation engines that generate statistically plausible but potentially fabricated outputs. His central thesis: these systems will profoundly reshape society not because they are intelligent, but because they are convincing liars.
Key Insights
What AI Really Is:
- LLMs are "giant piles of linear algebra" that predict statistically likely token completions
- They do not learn over time, do not remember, and do not understand
- They are improv machines that "yes-and" whatever input they receive
- They lie constantly — about operating systems, radiation safety, and the news
The Reality Fanfic Problem:
- LLMs treat sarcasm, fantasy, and fiction credulously
- They tell people to put glue on pizza (Google AI Overview incident)
- Humans are poor at distinguishing between statistically plausible outputs and truth
- LLMs hallucinate quotes and fabricate entire articles
Why "Lie" Is the Right Word:
Aphyr deliberately uses "lie" rather than "hallucinate": "Obviously LLMs are not conscious, and have no intention of doing anything. But unconscious, complex systems lie to us all the time. Governments and corporations can lie. Television programs can lie." The key distinction is that these are complex sociotechnical artifacts whose outputs can be false, regardless of intent.
Essay Structure
The work spans 10 sections: Introduction, Dynamics, Culture, Information Ecology, Annoyances, Psychological Hazards, Safety, Work, New Roles for Humans, and Where Do We Go From Here — available as a PDF and EPUB.
Why This Matters
Coming from someone who built their reputation on rigorously testing distributed systems claims, this essay carries unusual weight. Aphyr approach combines technical depth with accessible writing, making complex ML issues understandable to a broad audience.
Source: aphyr.com — 567 points, 557 comments on HN