The Waymo effect: how AI is quietly making research less collaborative
AI tools are seen by some as accelerating a shift from messy, time-consuming human collaboration toward frictionless, individual work with chatbots and agents, raising concerns about lost serendipity, weaker research culture, and erosion of deep expertise. Others counter that AI simply amplifies existing patterns—like overconfidence outside one’s domain or preference for convenience—and can democratize access to knowledge when used carefully. A large part of the exchange centers on whether AI‑generated writing should be called out or filtered, and how reliance on LLMs is already changing workplace dynamics, academic practice, and everyday social interaction.
AI, Waymo, and human interaction
- Many agree: self‑driving rides reduce small talk and social “friction”; some love this (no chit‑chat, no music conflicts, perceived safety, reliability, no tipping).
- Others argue those “inconvenient” interactions are part of social life and community; losing them deepens isolation and bubbles.
- Several think using drivers as the “last stranger” outside one’s bubble is overstated; they get diverse interaction from many other contexts.
Collaboration, research, and AI tools
- Commenters see a parallel with open source: output volume rising while community engagement declines.
- Some predict collaboration will have to be intentional; incidental coordination (hallway chats, drivers, “third places”) is being hollowed out by frictionless tech.
- Others argue if collaboration is genuinely valuable, people will keep doing it; tech just changes where and how, not whether.
LLMs, expertise, and overconfidence
- Strong theme: AI makes it easy for non‑experts to sound authoritative in domains they don’t understand, eroding respect for domain expertise.
- Reports of people arguing technical/financial/medical points using LLM‑generated fragments, then “checking” experts against the model.
- Some see LLMs as useful for probing doubts and learning faster; others say they mostly reinforce user bias and produce shallow or wrong analysis unless heavily verified.
- There’s concern that younger or newer practitioners may learn to manage agents and output, but not to develop deep understanding.
Debate over AI‑generated articles and detectors
- Large subthread insists the linked essay is LLM‑written, citing stylistic “tells” (stock constructions, repeated tropes, adverbs like “quietly”).
- Others push back: humans have always written that way; accusing without strong evidence risks chilling genuine writing.
- Pangram, an AI‑text classifier, is frequently cited as “actually good,” but some worry about false positives and over‑reliance.
- One camp treats “AI slop” labeling as useful spam‑like filtering; another calls it a form of witch‑hunt that can unfairly dismiss real work.
Broader views on AI’s social role
- Some see AI as individual empowerment that threatens institutions and gatekeeping; others see it as commodifying shared knowledge rather than democratizing it.
- Multiple commenters worry about “zero‑friction” culture: optimizing for convenience undermines long‑term benefits of effort, community, and serendipity.