New Study: Waymo is reducing serious crashes and making streets safer

Insurance, Liability & Economics

• Commenters expect liability costs to fall with safer AVs, but note large operators often self‑insure fleets rather than buy consumer policies.
• Self‑insurance is common in health and fleet contexts; insurers may lack actuarial data for AVs, leading to high premiums until more history exists.
• Some argue AV companies can be fined and held liable, but others worry they’ll be put in a different legal category than individual drivers.

Human Oversight & Scalability

• Debate over how much human “remote assistance” is needed: claims range from “non‑scalable 20:1 vehicles:staff” to “rare interventions, trending down.”
• Assistance is described as high‑level guidance (e.g., drawing a path around a fire truck), not remote driving. Critics say this still reflects AI’s limits in unstructured situations.

Safety Data, Statistics & Methodology

• Supporters emphasize tens of millions of public miles, strong reductions in serious and intersection crashes, and peer‑reviewed and third‑party‑involved studies.
• Skeptics question:
  • Sample size vs. national 3.2T vehicle miles.
  • Selection bias (limited cities, good weather, well‑maintained SUVs).
  • Comparisons to all humans vs. sober, middle‑aged, professional drivers.
  • Industry‑authored research and lack of fully independent audits.
• Some note prior Cruise data showed worse injury rates than humans, arguing that precise per‑mile numbers and unbiased baselines are essential.
• There’s discussion of benchmarking against an “ideal attentive human” model and of airplane‑style root‑cause investigations.

Experiences on the Street: Pedestrians, Cyclists & Traffic

• Many SF cyclists and runners report Waymos as consistently law‑abiding, predictable, and much safer to be around than human drivers.
• Others recount problematic maneuvers: creeping into crosswalks, entering bike lanes too closely, odd behavior around construction, and a few red‑light or strange right‑turn incidents.
• Several note “platoon pacing”: AVs strictly obeying speed limits can calm traffic and reduce speeding feedback loops. Others argue some speeding is beneficial for road throughput.

Edge Cases & Driving Behavior

• Concern about rare “split‑second swerve vs. brake” scenarios; some think human instinct is better, others point to faster reaction and broader awareness from sensors.
• Snow/ice performance is flagged as largely unproven and outside current domains. Some expect AVs simply to avoid operating in such conditions.

Privacy, Surveillance & Market Power

• Significant unease about fleets of sensor‑rich vehicles recording everyone in public space, with unclear retention and secondary uses.
• Fears of future monopolistic control, price hikes, and government surveillance coexist with enthusiasm for large safety gains.
• A few participants suspect astroturfing or excessive pro‑Waymo sentiment and question HN’s neutrality.