How Tesla is proving doubters right on why its robotaxi service cannot scale

Original Article ↗ Hacker News Discussion ↗

Broken Link and What “Robotaxi” Is Today

  • AOL link was broken; discussion points to a Fortune piece about Tesla’s Austin pilot.
  • Commenters stress Tesla’s “robotaxi” currently has a safety driver in every car plus remote teleoperators; it’s framed as a regular taxi service, not true driverless like Waymo’s mature operations.
  • Some note all robotaxi programs (Waymo, Cruise) started with safety drivers, but others point out Tesla has claimed a big head start and still lags.

Vision-Only vs LiDAR/Radar: Core Technical Dispute

  • Large subthread debates Tesla’s cameras‑only FSD versus competitors’ LiDAR+radar+camera stacks.
  • Critics: “no LIDAR no ride”; vision-only is fragile with glare, fog, dust, unusual objects, and non-standard pedestrians. Tesla is accused of prioritizing cost and simplicity over safety.
  • Supporters: modern FSD uses an end‑to‑end neural net with an internal world model; the dashboard visualization is not the driving model. Extra sensors add complexity and validation burden; a human-like vision stack plus huge data may be enough.
  • Others argue additional sensors are cheap relative to crashing, and industry practice in safety‑critical systems is to favor diverse sensor fusion.

Safety, Incidents, and Opaque Metrics

  • Examples cited of Teslas driving toward trains, misreading motorcycles, confusing freight trains, and needing frequent interventions; one rider’s near‑train incident in Austin is widely referenced.
  • Waymo is repeatedly praised by riders for smooth handling of odd situations and having no at‑fault injury crashes so far; some fear Tesla’s failures will taint the whole robotaxi sector.
  • Fierce argument over Tesla safety stats: fans claim FSD/Autopilot is much safer per mile than humans; skeptics say Tesla’s methodology is incomparable to Waymo’s more transparent reporting and excludes many incidents.
  • NHTSA’s rule that any crash within 30 seconds of ADAS disengagement counts as “engaged” is mentioned; Tesla is also accused of trying to block public release of detailed crash data.

Scalability, Economics, and Strategy

  • One camp: Waymo’s geofenced, HD‑mapped, multi‑sensor level‑4 model is safer but expensive and slower to deploy; Tesla’s vision‑only, map‑light approach is the only one that can truly scale “anywhere a human can drive.”
  • Opposing camp: unconstrained operational domain is “one of the stupidest ideas” in AV; real‑world performance (critical disengagement ~ every few hundred miles) shows Tesla is far from unsupervised use.
  • Business debate: Tesla’s early removal of radar/LiDAR is seen by some as a brilliant cost and data‑scale play, by others as premature optimization that now traps them technologically and legally.

Robotaxis vs Public Transit and Urban Capacity

  • Many argue even perfect robotaxis cannot solve congestion; thousands of 1–2 person cars will always move fewer people than buses, trams, or subways.
  • Others counter that US politics and timelines make large‑scale transit expansion unrealistic, so improving car‑based mobility (including AVs) is the only near‑term path.
  • Side debate over public transport quality: European and Asian systems are held up as proof it can work; US systems are portrayed as unsafe, dirty, and underfunded, driving demand for private or robotaxis.

Musk’s Credibility and Behavior

  • Musk’s meme‑shaped Austin service map, 4.20/6.90 pricing jokes, and long history of overpromising FSD “next year” are widely cited as reasons to distrust his timelines and technical claims (e.g., “photon counting” cameras).
  • Some still argue his track record with rockets and EVs means betting against him is unwise; others say those successes coexist with clear duds and chronic exaggeration.