Latest / Elon Musk Podcast / Waymo now offers driverless freeway rides in SF, LA, and Phoenix

Transcript

Waymo now offers driverless freeway rides in SF, LA, and Phoenix · Elon Musk Podcast

• 0:03
  Your ultimate authority for daily Elon Musk news.
• 0:06
  Exploring the world's biggest ideas with your host Will
• 0:09
  Walden. There's something new every day.
• 0:18
  So Waymo just switched on fully driverless freeway rides for
• 0:23
  paying passengers in San Francisco, Los Angeles and
• 0:26
  Phoenix. In the expansion connects SF to
• 0:30
  San Jose with curbside access and Minita San Jose
• 0:33
  International Airport. Now, is this the moment
• 0:36
  robotaxis move from the city streets to true region to region
• 0:40
  transportation that everyday riders will actually use?
• 0:44
  Tesla does not do this yet. And freeways are the backbone of
• 0:47
  how people travel in these metro areas.
• 0:50
  So letting the service take the fastest route changes the value
• 0:54
  of a robo taxi trip. Riders who were limited to
• 0:57
  surface streets can now do airport runs, crosstown commutes
• 1:02
  and city to suburb hops without a human being behind the wheel.
• 1:06
  And the company says this rollout follows a steady safety
• 1:10
  record in city driving. And it is extending that
• 1:12
  approach to higher speeds and more complex merges.
• 1:19
  Now freeway access is the unlock that turns the novelty into a
• 1:22
  network. The post break picture here is
• 1:25
  simple. Then it gets nuanced.
• 1:27
  First the capability. Vehicles can enter and exit
• 1:31
  freeways, keep pace with traffic, and choose the route
• 1:34
  that actually gets people there faster when the limited access
• 1:38
  Rd. is best. Then this step arrives after
• 1:43
  years of restricting service zones and gradual expansions in
• 1:45
  the Bay Area and Phoenix, and after months of controlled roll
• 1:49
  outs in Los Angeles neighborhoods.
• 1:52
  What matters now is how the service scales across corridors
• 1:56
  that millions of people rely on, and how Tesla responds.
• 2:01
  And in practical terms, riders in the Bay Area can book trips
• 2:05
  that include freeway segments between San Francisco and San
• 2:08
  Jose, and they can get dropped curbside at the San Jose
• 2:11
  airport. That removes one of the most
• 2:13
  obvious gaps in earlier service maps where airport access and
• 2:17
  cross peninsula travel forced route compromises.
• 2:21
  LA riders should see more direct routing between neighborhoods
• 2:25
  split by wide interstates, and Phoenix riders already familiar
• 2:28
  with driverless trips can now use freeways when that is
• 2:32
  objectively the quickest option. Now the company positions this
• 2:38
  as a proof of generalization. The driver handled dense
• 2:42
  downtown's first, then complex suburban arterioles, and now it
• 2:48
  steps into sustained high speed traffic with more lane changes,
• 2:53
  merges, and incident handling. The claim is not that freeways
• 2:58
  are easier or safer by default, but that the system's perception
• 3:02
  prediction in planning stack has matured enough to operate at
• 3:07
  freeway speeds while maintaining its conservative margins.
• 3:11
  You know the internal message is expansion methodically, not a
• 3:16
  single headline moment. They're not looking to grab
• 3:20
  attention here. They are working to make a robo
• 3:25
  taxi that can drive you from network to network, from city to
• 3:30
  city, from state to state, all across the United States.
• 3:35
  Now, this roll out also relies on clear playbooks for awkward
• 3:39
  edge cases. And if a vehicle defects on the
• 3:43
  freeway, it'll pull over safely. A response protocol in
• 3:48
  coordination with Highway Patrol will then avoid secondary
• 3:52
  incidents. Those procedures exist in city
• 3:55
  service, where cars can stop at curbs with freeway shoulders.
• 3:59
  Ramps and interchanges demand different timing and
• 4:02
  communication. Now.
• 4:04
  The team says it has worked with authorities on the scenarios,
• 4:08
  and that collaboration will matter if the service is to keep
• 4:12
  public confidence as miles accumulate.
• 4:15
  Now scale is the other story, and the numbers point to a
• 4:18
  service that has moved beyond pilots.
• 4:21
  Public reporting from recent weeks describes more than 1500
• 4:24
  vehicles in the fleet, with steady weekly ride counts in the
• 4:28
  hundreds of thousands. And the mix is still anchored by
• 4:31
  electric crossovers equipped with prioritary lidar, cameras,
• 4:36
  radar, and onboard compute designed for redundancy.
• 4:40
  Now, hardware matters here because freeway driving brings
• 4:44
  sustained vibration, higher closing speeds, and the need to
• 4:48
  read both lane level markings and long range signals well
• 4:52
  enough to plan several 100 meters ahead.
• 4:56
  And competitors know freeway credibility is the threshold
• 4:59
  that you have to cross to Uber plans to treat field driverless
• 5:03
  taxis in San Francisco use and lucid gravity SU VS with
• 5:09
  autonomous technology from Nero moving beyond a booking
• 5:13
  partnership into direct competition on those same
• 5:15
  streets. And the interim goal is to build
• 5:17
  toward more than 100 vehicles during the ramp.
• 5:21
  And if schedule holds, Bay Area riders will compare 2 driverless
• 5:24
  services next year and also Tesla.
• 5:28
  And that will be coming up. And the deciding factors will be
• 5:31
  reliability, pick up, latency, trip time and, of course, price.
• 5:37
  Tesla is pushing a different path, though that still requires
• 5:40
  an attentive human ready to steer or brake at any moment for
• 5:44
  now. And the company markets the
• 5:46
  experience with a robot taxi label in its app.
• 5:49
  But the underlying system today is supervised driver assistance,
• 5:52
  not a completely driverless service.
• 5:55
  And that distinction matters on a highway, because a service
• 5:58
  that requires human fall back cannot offer the same value
• 6:01
  proposition as a fully unattended ride, and it cannot
• 6:04
  claim the safety or availability profile of a Level 4 fleet.
• 6:10
  Now Amazon is in this, too. Zoox is expanding its own
• 6:14
  footprint with free rides around the Las Vegas Strip, and those
• 6:18
  deployments will keep pressure on all players to show smooth
• 6:21
  autonomy and complex traffic. But the bar for freeways is
• 6:25
  completely different. It is not enough to handle
• 6:27
  predictable loops. A driverless service has to
• 6:30
  merge in heavy traffic, navigate unpredictable human behavior at
• 6:34
  on ramps and off ramps, and keep situational awareness across
• 6:38
  multiple lanes at 55 to 70 mph. That is why this particular
• 6:44
  expansion is seen as a marker of technical maturity rather than a
• 6:48
  simple map update. And the policy backdrop is
• 6:52
  moving in parallel. City and state officials are
• 6:54
  trying to balance the promise of fewer crashes and lower
• 6:56
  emissions with reasonable guardrails for new risks.
• 6:59
  Some local proposals would require a human operator and any
• 7:03
  autonomous vehicle, which would collapse the economics that make
• 7:06
  driverless service compelling. Others focus on incident
• 7:09
  reporting, data sharing and coordination with first
• 7:12
  responders, which can improve transparency without freezing
• 7:16
  the technology in place. How those choices land will
• 7:19
  shape where and how quickly freeway service rolls out next.
• 7:23
  Safety investigations are part of the learning curve, too, and
• 7:26
  everyone in the sector is dealing with them.
• 7:28
  Reports or unexpected driving behavior have drawn federal
• 7:32
  attention, and the result is a more formal feedback loop
• 7:35
  between regulators and companies.
• 7:37
  The outcome the public should watch for is not the absence of
• 7:40
  scrutiny, but the presence of clear fixes, lower incident
• 7:44
  rates over time, and better explanations of how the systems
• 7:48
  are tested before features turn on.
• 7:50
  For the general public and for riders, the litmus test is
• 7:54
  simple. Does the car show up when the
• 7:56
  app says it will pick a route that makes sense and complete
• 7:59
  the trip without odd pauses or last minute reroutes?
• 8:02
  Freeway access improves all three because a service can't
• 8:05
  avoid fragile Surface St. shortcuts and take the obvious,
• 8:09
  faster straight line path. It also enables more convincing
• 8:13
  airport trips and cross metro rides that felt unnatural when
• 8:16
  the car refused to enter on the highway and the business.
• 8:22
  If a driver of this fleet can handle freeways reliability, it
• 8:26
  can spread fixed costs across more miles, use vehicles more
• 8:29
  hours per day, and price trips more competitively against ride
• 8:33
  hail with human drivers. That is where the competitive
• 8:36
  dynamic tightens. The player that proves it can do
• 8:39
  this safely at scale with predictable margins will set the
• 8:43
  preference point for the rest of the industry, and they may find
• 8:46
  itself defining the regulatory expectations others must meet
• 8:52
  now. The engineering leap took years
• 8:54
  because the problem seems simple and it's not.
• 8:58
  The freeways look structured, but the emergent behavior is
• 9:01
  messy. People are weird.
• 9:03
  On the highway. A system has to predict whether
• 9:05
  a driver 2 lanes over will dive across to make a really quick
• 9:09
  exit and cut a bunch of people off, whether a stop vehicle on
• 9:12
  the shoulder hides A pedestrian about to step into traffic, and
• 9:15
  whether an object ahead is safe to straddle or requires a lane
• 9:18
  change right now. And that demands perception
• 9:21
  range, multi agent prediction and a planner that chooses
• 9:25
  smooth human like decisions without taking risks a
• 9:28
  professional safety driver would reject.
• 9:30
  And they have to talk about different things.
• 9:33
  Now you can hear how the rollout is described.
• 9:36
  It's not framed as a bold stunt or a sudden leap forward.
• 9:40
  It's just calm. It's presented as the next
• 9:44
  capability enabled by a driver that has already survived hard
• 9:48
  miles in the city. Learn to deal with cut insurance
• 9:51
  and builds up the operational protocols to handle the
• 9:54
  exceptions. Now this company is connected
• 9:56
  the dots between city service areas in the corridors that make
• 10:00
  those areas feel like one region.
• 10:03
  City to city, the competitive responses will keep coming, too.
• 10:08
  Uber wants to become both the marketplace for other robotaxi
• 10:11
  fleets and a direct operator. Tesla's pushing toward
• 10:14
  unsupervised capability from a supervised base.
• 10:18
  Zoox is packaging autonomy with purpose built vehicles, and each
• 10:23
  path is coherent. But the freeway litmus test will
• 10:25
  expose differences in system maturity, safety, performance,
• 10:29
  and user experience quickly because riders do not forgive
• 10:34
  weirdness at 70 mph. Now the next meaningful
• 10:39
  questions are the ones riders will ask without thinking about
• 10:42
  autonomy at all. How long a time until a curbside
• 10:45
  pickup is available at more airports?
• 10:48
  When will late night rides feel just as smooth as daytime trips?
• 10:51
  And how quickly do service maps fill the gaps between
• 10:54
  neighborhoods that still require awkward detours?
• 10:57
  The answers will come from miles, hundreds of thousands of
• 11:02
  miles of usage. If freeway trips become routine,
• 11:05
  volume will rise and the service will feel less like a pilot and
• 11:09
  more like transport infrastructure you can rely on
• 11:13
  any time of the day. Hey, thank you so much for
• 11:17
  listening today. I really do appreciate your
• 11:19
  support. If you could take a second and
• 11:21
  hit this subscribe or the follow button on whatever podcast
• 11:24
  platform that you're listening on.
• 11:25
  Right now I. Greatly appreciate it, it helps
• 11:28
  out the show tremendously and you'll never miss an episode and
• 11:32
  each episode is about 10 minutes or less to get you.
• 11:35
  Caught up quickly and. Please, if you want to support
• 11:38
  the show even more, go to. Patreon dot.
• 11:40
  Com slash. Stage 0.
• 11:43
  And please take care of yourselves and each other, and
• 11:46
  I'll see you tomorrow.