Waymo: The future of (not) driving
Waymo is attempting to improve transportation using autonomous vehicles. As it is preparing to launch a ride-hailing service with these cars, there are several challenges it faces – from perceived safety issues to long-term transportation policy around the world.
Waymo LLC, a subsidiary of Alphabet Inc. is building technology that makes cars fully autonomous and is currently in the process of launching a ride-hailing service with a fleet of these cars. None of this would have been possible without the application of machine learning (ML) techniques, specifically, deep learning. At Waymo, ML isn’t merely a tool used to optimize a process or provide better insights into data; without it, the organization wouldn’t exist. Waymo is arguably the leader in autonomous driving technology, as measured by the total number of miles driven or simulated [1]. With all this technology at its disposal, and the financial backing of an organization as large as Alphabet, is there any scenario other than one in which Waymo leads the world in the future of transportation and mobility?
There’s no arguing that ML techniques form the cornerstone of Waymo’s business. The technology that enables self-driving vehicles relies on sensors to identify the surroundings of the car and makes decisions which mirror and potentially improve on human driving. While sensors to assist drivers park, navigate etc. have been around for a while, the critical component here is the algorithm that analyzes all this data and then controls the car.
Image source: Engadget [1]
Despite all the simulated and real-world testing, their algorithm is not perfect and is subject to external factors like other drivers on the road. Waymo’s cars have been involved in multiple crashes but it appears that most of them were caused by human violations [2]. However, this is proving to be an issue for the company in the short term as it is preparing to launch a fleet of autonomous taxis in Arizona. Although accidents involving autonomous vehicles (AV) are far fewer than those caused by humans [3], there’s growing discomfort about the safety of these vehicles. To this end, the company has started an “Early Rider” program [4] which allows interested riders to use the service (for free). In order to improve rider safety, every car will have a safety driver at all times. In addition to introducing a new product to customers, Waymo can also gather data about customers’ perceptions of and reactions to the product and this is more data for their algorithm to learn and improve the ride experience. I think investing upfront in educating customers about the benefits of AVs will help build a loyal customer base and considering that this is still a novel product, customer interest is still high and a directed marketing campaign can help build brand loyalty.
Waymo believes that AVs will make transportation inherently safer by removing the common mistakes that humans make on the road [5]. However, to be completely safe, there needs to be a critical mass of AVs on the road, but this is a long-term continuous issue for the company (and industry) as legislation supporting AVs is almost nonexistent. Waymo, through its parent company, Alphabet is investing in legislative efforts [6] to educate policymakers about AVs and associated benefits in order to direct regulations and infrastructure investments to support the industry. In order to truly be a partner to the public transportation systems, Waymo is also attempting to factor existing infrastructure into its models to best serve customers [7], which will help the organization build a symbiotic long-term relationship with multiple stakeholders and improve its public perception. While this is a continuous process for the company, I think Waymo can demonstrate its commitment to its mission of improving transportation by assisting cities in emerging economies, where it wants to expand, build integrated transportation systems.
I have little doubt in my mind that AVs will be a staple of our transportation systems in the future. However, the transition from existing systems will differ drastically across the world and the implications will be widespread, and I think the manufacturers will have to play an active role beyond just product development. While I believe there are definite tangible benefits of an AV versus a regular vehicle, there are several questions to which I lack answers. For example, if an AV is involved in an accident, should the manufacturer be liable? Are the net benefits of AVs sufficient to negate the effects of job losses in the transportation industry? Given the requirements of AVs to be safe and successful, will everyone be limited to only AVs in the future and how will the regulations impact competition in the industry?
(725 words)
Endnotes:
- Andrew J Hawkins, “Inside Waymo’s Strategy to Grow the Best Brains for Self-Driving Cars”, theverge.com, May 9, 2018, https://www.theverge.com/2018/5/9/17307156/google-waymo-driverless-cars-deep-learning-neural-net-interview, accessed November 2018
- Alison Griswald, “Waymo’s self-driving car crashed because its human driver fell asleep at the wheel”, qz.com, October 2, 2018, https://qz.com/1410928/waymos-self-driving-car-crashed-because-its-human-driver-fell-asleep/, accessed November 2018
- Aarian Marshall and Alex Davies, “Waymo’s self-driving car crash in Arizona revives tough questions”, wired.com, May 4, 2018, https://www.wired.com/story/waymo-crash-self-driving-google-arizona/, accessed November 2018
- Waymo LLC, “Early Rider Program” https://waymo.com/apply/faq/, accessed November 2018
- Waymo LLC, “Technology”, https://waymo.com/tech/, accessed November 2018
- Tony Romm, “The super-secret Google X lab has hired its own lobbyists in the nation’s capital, recode.net, October 19, 2017, ”https://www.recode.net/2017/10/19/16506030/google-x-money-alphabet-lobby-lobbying-congress, accessed November 2018
- Abner Li, “Waymo explores integrating self-driving cars w/ public transportation for the ‘last mile’, 9to5google.com, July 31, 2018, https://9to5google.com/2018/07/31/waymo-self-driving-public-transportion/, accessed November 2018
Image sources:
- Richard Lawler, “Waymo’s 360-degree demo ride shows what self-driving cars ‘see'”, engadget.com, February 28, 2018, https://www.engadget.com/2018/02/28/waymo-360-degree-vr-self-driving-car-demo/, accessed November 2018
I share the belief that AV will be a staple of our transportation system, and very probably dominate it. My biggest concern is about the capacity of the algorithm to make tough ethical challenges such as for example taking a self-inflicting damage driving decision to avoid crushing and killing more people (choosing the “lesser” evil/bad outcome). And although this can get subjective (what is the lesser bad outcome), the image recognition functionalities should be working at 100% with absolutely no chance of mistakes (google in the near past has been having trouble with image recognition https://www.theregister.co.uk/2017/04/19/cloud_vision_api_defeated_by_noise/). That said, I think one of the benefits of the model is that it will increase car utilization, thus reducing the number of cars on the streets and the levels of pollutions caused by cars.
I, too, have little doubt that AVs will be a staple of our transportation systems in the future. From individual vehicles to public transportation, I truly believe autonomous will be the norm. However, I agree with you that the transition from existing systems will differ drastically across the world and the implications will be widespread. That said, transportation is inherently from point A to point B. What happens when one city advances much faster than the other? As you said, to be completely safe, “there needs to be a critical mass of AVs on the road”. Also, as we’ve seen in the mobile market, Androids and Apple operating systems are inherently not compatible. From charging wires to applications that we download to use, there exist different versions for each. What happens when not just Waymo but other AV makers hit the road? How will these cars drive next to each other? Will it be safe for the riders?
Like Tomas, I’m struck by the morally consequential decisions that will need to be built into some of these algorithms. Specifically, what will be the relationship between businesses creating these technologies and government regulators? Will government regulators focus only on (suboptimal, after-the-fact) outcomes, or will they become part of developing these algorithms and be a partner in navigating the ethical challenges that Tomas mentions above?
I think the ethical questions will likely fade away as we move from theoretical discussions to practical application. These vehicles will be designed such that they (almost) never get into accidents for which they are at fault and even if one becomes unavoidable, they will likely be designed chose a path all the way up to the very end that minimizes the chance of a collision. Another possible cause of an accident would some kind of system failure and in this case we may have relatively little control over how the vehicle responds.
You raise an interesting questions about a future in which only autonomous vehicles are allowed on the road. I personally don’t think this is likely for a few reasons. First, the current fleet of non-autonomous cars on the road will be around for a very long time, meaning that autonomous vehicles will be required to navigate around them and therefore a shared road should become a non-issue as AV manufacturers meet this requirement. Secondly, there is still a strong set of Americans who are passionate about there vehicles (including car dealers who make up a very powerful lobby). Because of this, I think taking away car enthusiasts manually operated cars will be similar to trying to take guns from gun owners.
Thanks for writing this! In addition to what you’ve written in the article, I’d like to pose additional benefits of autonomous vehicles for urban dwellers. If the majority of vehicles are autonomous and ride sharing, this will likely eliminate the need for on street parking as the vehicles will not be owned by individuals but by companies like Waymo and can be parked near or outside the city’s limits. That frees up a lot of space in the public right-of-way for parks, bike lanes, scooter lanes, and more!
I am personally very excited to see autonomous cars on the road because I don’t know how to drive! As we have seen with the accident in Arizona that you mentioned, scrutiny on the safety rate of autonomous driving will be even more severe despite the well-known dangers of human driving; it’s difficult to place the blame on a human over a machine, even if it is the human that violated traffic rules. Is the only way to convince regulators of its safety continued testing?
I am surprised to see the progress that autonomous driving is seeing on the regulatory front, being allowed recently to drive without a monitoring human passenger — we may see widespread use of this technology sooner than we think, especially as more people are using ride-hailing services. I agree with you that a robust regulatory infrastructure must be established before this happens, though, with accountability determined for all possible outcomes.
I think your points are not only valid for autonomous driving, but also but also valid for the future of automation. I think we are still far away from having a network of AV dominated structure. Apart from the technological advancements necessary, I believe the concept of liability is not totally applicable in this context – I think if we reach to a point where having AVs is widespread, that level of technology and the intelligence of the cars makes it really difficult for the liability to be associated with the manufacturer. The car would became almost a human in terms of thinking capability, and how could us associate anything after that with the manufacturer? I hope we would have a better answer to this once we have some more visibility about the technology to be developed in the future.
It is interesting to think about how the rise of AVs will have ripple effects in other industries – what does a world with all AVs look like. The dynamics of several even outside of the traditional manufacturing companies will be dramatically changed, as will how we interact with cars generally. When car ownership drops dramatically as a result, how do manufacturers compete with each other. Who actually own cars in a world with all AV and who is accountable for when those cars get into some kind of accident: property rights and therefore liability are difficult to describe in a system where decisions are deterministic.
The driverless car revolution could come much sooner or much later than we think. It’s plus or minus decades. That’s important because it means a driverless car company would need to be able to survive long enough for this technology to catch on. Advantage anyone backed by Google (or other large profitable tech company with deep pockets), disadvantage anyone backed by a VC funding round. To me that is Waymo’s biggest competitive advantage in this space.
I do think it will take a while before we can expect an all-AV scenario, Brian, however I do think that we will get to a high level of adoption (~90%) by 2050. This is namely for the safety this technology brings. Although lacking in many ways, there are gun control regulations that reduce the prevalence in society; I think we can expect the same due to the risks to human life. Similar to guns, I think there will be dedicated spaces for car enthusiasts willing to pay for their opportunity to drive in public areas (and drive internal combustion cars as long as they’re willing to pay for the cost of emissions).
@ssingayapally did you encounter discussion on the need for Lidar technology and its role in providing data for Waymo’s ML algorithms? I know other companies such as Tesla don’t view this technology as necessary so am curious to learn more about why Waymo and others believe the extra visibility it provides is necessary / worth the extra cost and physical impact to vehicle design. Also, does the fact that Waymo is partnering with FCA and Jaguar to obtain vehicles as opposed to manufacturing themselves pose any risks to long term development of their product?
Thank you for the article, I really enjoyed it. It’s amazing what Waymo is achieving so fast.
Three challenges I have on mind when thinking about self driving cars:
1. Ethics – What should be the main role of the vehicle, to protect the driver or to minimize the number of potential deaths? If the second, who will want to get in a self driving car? If the first, will we be in an inefficient system? I find this web page from MIT a very fun way to think about the topic. http://moralmachine.mit.edu/
2. Cyber Security – How vulnerable are self driving cars to being hacked and used for crimes? We have already seen this in movies (Fast and Furious 8) and TV Shows (The Good Wife), but how much time has to pass until a case happens in real life, and what will be the impact of this on the legal framework?
3. Business model – At what point will cars switch from ownership to a pay-per-use model? We already have Uber, but Uber doesn’t own its cars, people do. So, will self driving cars be seen as a personal belonging that can be used by Waymo and Uber during idle time? Or will big funds buy the cars and offer them to Uber?
Thanks!
I fear that it will take at least 20 years before driverless cars (no assistance from humans) will be considered a normal means of transportation. As discussed in the article, there are major concerns surrounding the reliability of driverless cars and who will be at fault when accidents occurs (driverless car vs. human, driverless car vs. driverless car). The first time that a driverless car is involved in a fatal car accident, I predict that regulations and laws will tighten. How will driverless car manufacturer present there case in court? Would data logs be considered testimony, or even allowed, and how will the court interpret it? Subsequently, driverless car manufacturers will incur expenses deriving from law suits and the need to advertise to combat negative publicity. This will slow down the industry’s moment. I think the current state of the industry, where driverless cars have a driver as safety measure, is where the industry will stall.
“I think Waymo can demonstrate its commitment to its mission of improving transportation by assisting cities in emerging economies, where it wants to expand, build integrated transportation systems.” Couldn’t agree more with this point! As AV technology improvements hit a plateau, companies like Waymo and Tesla can no longer rely purely on product development to get to the safety levels needed for general public adoption. Just as roads and laws/regulations changed with the introduction of cars, so too will these aspects need to change with the introduction of autonomous vehicles. I think the sooner manufacturers recognize this, the sooner we’ll be able to see the next wave of improvements come online!