Creating an all-weather driver

When should an autonomous car stop driving?

• Several commenters wonder how the system decides it’s “too bad to drive,” noting humans are bad at this and often overestimate their own skill.
• Some fear a “liability-maximizing” dystopia where the car refuses to attempt escape from a storm; others say that’s preferable to overconfident systems crashing.
• There’s skepticism that tech can ever fully avoid risks like black ice; at best it can manage consequences better and maybe recover from spins with superhuman control.
• Some insist winter crashes are mostly “skill issues,” while others push back, saying some hills/ice conditions are effectively unmanageable.

Human habits and culture in bad weather

• People learn safe responses (pulling over in heavy rain, using hazards, avoiding known icy hills) mostly by experience and local lore, not formal training.
• Practices differ by region (Florida rain vs Texas hail vs European snow vs US Midwest/New England tolerance for snow on all-season tires).
• Debate on hazard lights: some think they should be used whenever going far below the limit; others say they’re for stationary/true hazards only.

Hardware: chains, tires, and traction

• Chains are common only in mountainous regions or where legally required; many US drivers use all-season tires year-round, even in serious winter.
• Some argue dedicated snow or 3PMSF all‑weather tires are vastly better; others say an AWD SUV on all‑seasons is “good enough” for most people.
• Automatic deployable chains exist on some emergency vehicles and school buses; even they can’t handle certain steep icy spots known only to locals.
• Commenters suggest fleets could swap tires seasonally based on forecasts. Cost and fuel‑economy pressures push manufacturers toward hard, efficiency‑optimized tires.

Interacting with police, workers, and ad‑hoc directors

• Waymo cars reportedly got stuck at a Los Angeles event, unable to interpret police hand‑waving at crossings.
• Many see informal human signaling (cops, road crews, random “volunteer” traffic directors) as one of the hardest remaining problems.
• Ideas: standardized machine‑readable signals/devices; authenticated override tools like “Waymo keys” for emergency services; or ultra‑cautious behavior around anything that looks like an emergency.
• Others doubt standardized gear will be reliably used, given real‑world variability and low‑bid contractors.

Sensors: cameras vs lidar/multi‑sensor

• One camp argues cameras alone are “in principle” sufficient since humans drive with vision; they expect AI vision to eventually match or exceed human ability.
• Another camp says current camera‑only systems are obviously underperforming (struggling even with wiper control), while lidar‑based fleets are already operating without in‑car safety drivers.
• Several note humans don’t drive on “vision only”: we use sound, vestibular sense, haptics, and adaptability, so cars may need richer sensor suites to truly match us.
• Some expect society to demand superhuman safety from machines, making multi‑sensor systems the likely long‑term standard.

Geography, testing, and “hard modes”

• Commenters highlight Upstate/Western New York (lake‑effect snow), Sierra Nevada passes, and Boston city driving as especially valuable or brutal test environments.
• There’s debate over how unique US winter culture is versus Europe, with regional variation inside the US emphasized (Buffalo vs warm‑climate cities).

Driving tests and navigation UX

• Detailed European-style driving exams (snow‑covered courses, strict parallel parking) are contrasted with comparatively simple US tests, raising questions about how “average driver skill” is defined.
• Some wish Google Maps incorporated more of Waymo/Street View’s understanding of complex intersections; others complain current lane and speed‑limit guidance is still unreliable.