I recently attended NXP's Silicon Valley event called NXPConnect. Kurt Sievers, the President of NXP, opened the show by pointing out it was their largest show ever, with over 1500 people attending, over 100 hours of technology lessons, and over 90 demos in the tech lab. These demos show everything from vacuum cleaners to autonomous cars, but with a theme running through that everything has sensors, is always connected, and is smart.

I actually went mainly to attend the two automotive panels, doing my homework before my annual pilgrimage to Ludwigsburg outside Stuttgart. I just got back yesterday so my posts from there will appear after the holidays. I'll cover the first panel today. Panels are often a little boring because everyone is in such strenuous agreement. This one was fascinating and full of well-informed opinions.

Noisy Electric Cars

Nothing to do with the panel directly, but to do with electric cars, is the fact that from today (July 1st 2019) all new model electric cars in Europe must be fitted with an acoustic vehicle alert system (Avas). This is a fancy name for a noise-making device when traveling below 20 kilometers per hour (or reversing). Note that this only applies to new models from today. New cars from existing model lines are exempt until 2021, when all new vehicles will require Avas.

In the US, all vehicles need to comply with similar regulations by September 2020, but 50% of electric vehicles must comply by September 2019. The speed limit below which this is required is higher, at 30 kilometers per hour (which is a pretty weird choice of units for the US, 18.6mph).

Obviously, one motivation for this is to help blind people. But it seems to me that what the regulation calls "auditorially challenged pedestrians" is more likely to be someone looking down at their smartphone and not paying attention.

Self-Driving Cars: What's the Payoff for Carmakers?

This panel consisted of moderator Sam Abuelsamid from Navigant Research, along with panelists:

• Alex Tan from NXP
• Vijay Albuquerque of Luminar
• Wesley Shao of BYTON
• Rachad Youssef of NIO

Vijay said in his opening remarks that safety is the big issue. The challenge is what safety is needed and how much does it cost. Currently, with the number of cars, safety doesn't scale economically, so we will need to decide what safety is required and then scale it with volume to get the cost under control. Currently, there are just hundreds of vehicles, not thousands, let alone hundreds of thousands.

Wesley agreed with that formulation: Safety first, then scale for cost. But another variable is time. If we can achieve safety for 10 years then there is no point in having the cost discussion. For now, he doesn't see the cost coming down and the market won't happen until it does. $10K, $20K, nobody will pay it. The car companies are all working on this technology since it's their survival. It's not yet about cost for them. How can they not work on it? That doesn't mean that they will bring it to consumers before they are ready—they can't come to market without making money.

We are all hoping the cost comes down and the safety catches up.

Rachad agreed that time is a factor, but so is pent-up demand for this safety and convenience feature. This is a pivotal moment in the industry, like when flat-panel TVs took over completely from CRTs, and the market was shaken up with new entrants.

As far as timing, I think it is up to us, this audience, to cordon off an area where we can be successful and then grow from there.

Rachad thinks it is market and geography dependent: driving around Shanghai is valuable, driving down 101 from SF to SJ is valuable. But obviously, they are very different. The challenge is to generalize in the context of safety. He feels that vehicles need to be several times safer than a human in order to launch. "We need a pragmatic approach as to how we roll out the technology though."

Sam picked up on that idea and asked everyone how much safer than a human does an automation system need to be before we deploy and start to get benefits?

Alex went first.

I talk to a lot of car manufacturers and also people like Uber, etc. Generally the expectation is that it has to be considerably safer, several times safer. “Overall this has saved lives” is not that compelling versus anecdotes like a car driving into a truck. As a shared service, its more blurred since you are not talking whole risk.

Vijay told us to look at the number of interventions per mile driven (an intervention is when a human driver has to take over).

We are a long way from a human and we are looking at a couple of decades if we only improve at the current rate. So the technology has to evolve. You need 99 and ten 9s after that. On geofenced freeways it is much easier to target. In Phoenix, they have wide streets, well-behaved pedestrians, and good weather. Somewhere like San Francisco is much more difficult.

Wesley actually picked a number: ten times. "I think it needs to be ten times before the general public will accept...just a feeling from the psychology. Any accident like the Uber accident is going to be viewed as an issue for the whole industry, not just for Uber. I don't think it will be decades. But ten times."

Rashant agreed and said that within an order of magnitude is the right target given human psychology.

Two more factors are how do you prove it, and also how does it feel. Do users feel safer than driven by a human being? That is needed, too.

Sam moved on to return-on-investment. "Given this downbeat assessment, and the tens of billions of dollars being spent, is this ever going to pay off?"

Rashid believes. He acknowledged that for some time investors will support you selling at a loss, but we need to get a product to market that is compelling. "But today startups are truly being fueled by investors. It's still early."

Wesley reminded everyone that:

the future is autonomous driving and so eventually it will pay off. The automakers are slightly better off since they have an existing cash flow. But it's going to be a long process. It’s going to take many years. Optimist 2025, pessimist 2030. Nothing will pay off before then.

Vijay is optimistic. "I wouldn't be sitting here if I thought it was decades. 2025-2030 is realistic." He also pointed to some other important trends:

What ridesharing has proved is that there is a large number of users who will set aside their cars and use ride-share. I live in SF and I seldom drive a car there, it’s just not convenient. Look at the millennium generation and they are just not that interested in owning cars. It is encouraging that there are so many business models.

Alex is in no doubt that the money will be paid back eventually. It also goes beyond cars into things like drones and other devices that also need safety. "But there will be winners and losers over the next 10-15 years and some will be big. "Electrification and self-driving are both big challenges. Maybe one of the guys on this stage will one day be as big as GM or Toyota."

Sam wondered how this will impact the automakers. "Are they going to be left on the side becoming essentially tier-1 suppliers? How will this play out?"

Rashid emphasized that the car market is changing and there will be different ownership models. For instance, in China, you can purchase the car and lease the battery. "There's no one-size-fits-all for mobility."

Wesley thought that you can divide the market into China and the rest of the world.

Using self-driving robotaxis is a way for new energy [that's the terminology used in China for electric traction vehicles] to catch up with traditional. New entrants can be more advanced, and can catch up compared to the 40 different brands of Ford. But before 2025, robotaxi won't be financially viable since they will still need safety drivers. These are not normal drivers, chances are they are engineers. It's $20-30K of equipment per vehicle in 2023, plus the human side of the business. But there are benefits anyway to collect the data. You can even deploy without human drivers at airports and university campus maybe.

Vijay is also in agreement that the next 5-10 years is a period of development and learning on both the hardware and software side. "You have to get to where the costs are sustainable or you will never get to scale. That goes for both robotaxis and the carmakers."

Alex looked at the financials and the risks for traditional OEMs (carmakers):

At least in urban areas with younger demographics could see a scenario where rideshares offer self-driving cars, and if it takes too long from traditional OEMs they could see themselves fighting to be a supplier at all. There is a different ROI for a fleet for car rental versus private ownership. This is the challenge for car manufacturers. After all, you all have a smartphone in your pocket, but it's not that long ago they started, so changes can happen faster than you expect.

This is one reason it is so revolutionary. Unlike some other features leading up to it like cruise control, the value is no longer the most to the driver (who is usually owner). Now you don’t need a driver. Children and elderly have the same mobility as someone who can purchase a vehicle does today. The biggest value is not to people buying luxury vehicles but to the people who couldn’t afford a vehicle or can’t drive one. So will have a big impact on how these cars are deployed.

Sam opened the questioning to the audience. The first question was very direct: "When will level 5 arrive?"

Rachad said it's just an academic concept. If we have level 4 that satisfies the case people care about that is key. The rest is a gradual evolution. I’m on board with 10+ years. But we can get to very useful applications much earlier.

Alex said you might get level 5 before 2030 in geoboxed areas or in a constrained environment like a theme park. "For sure, some manufacturers will try out the concept."

The next audience question was whether it makes sense to try and amortize some of the development by advanced level 2 and higher level vehicles.

Vijay said that level 2 is already deployed.

Levels 3 and 4 can be deployed in regular cars. Deployment stages are evolutionary, will give us a good understanding of what is needed, and get public used to it.

The final question of the morning was about liability. What is the direction of liability in an autonomous rideshare? Who is responsible?

Wesley took the question. "What car is Google using in Arizona? Anyone know? If you ask on the street, people don't know. Google is going to pay because it has their logo on it. "When an accident happens, it's whose logo is on the car, influenced by local laws. There are some differences in China, but it is basically the OEM's issue. If it says Uber, then it is their responsibility. It's murky, but whose logo is the big thing."

Sign up for Sunday Brunch, the weekly Breakfast Bytes email.