There is Elon Musk vs. everyone else in complete driver without a car

At Tesla's "Autonomy Day" event for investors on Monday, Elon Musk was full of rubbish for his competitors and their technology.

LIDAR, the light beam sensor that virtually everyone sees as an important ingredient for self-driving cars, is "an idiot's errand," according to Musk. In addition, he continued, "Anyone who trusts LIDAR is doomed. Judged. Expensive sensors that are unnecessary. It's like having a whole bunch of expensive attachments … you'll see."

The simulation programs use to run "His virtual self-driving cars through millions of" edge "scenarios to raise billions of running miles are good, but not good enough," said Musk. "We also have a pretty good simulation, but it not only takes a long haul of weird things happening in the real world."

High precision GPS maps for self-driving cars are a "very bad idea", according to Musk, resulting in a "system [that] becomes extremely brittle" by being too dependent and unable to adapt.

Tesla will also have its autonomous level 5 vehicles that can run anywhere, under any conditions, ready to go by 2020, according to Musk. No geographical limitations. And not just a handful of cars, either. "A year from now we have over a million cars with full self-purchase, software, everything," he said.

For those who have seen Musk, make it divination for over a decade, Monday's event was not & # 39; t all that surprising. After all, it is the guy who tried to make his factory a "foreign dreadnought" of car-building robots. But to what extent he seemed to enjoy knocking down the key pillars of the self-propelled industry, it was truly astonishing, and many in the industry are confusing about how he can deliver his promises.

Musk has long argued that Tesla's self-propelled benefit comes from having a large car fleet ̵[ads1]1; around 425,000 – already on the road. These cars address situations and provide training data to improve the neural networks needed for self-driving cars. The company's approach to autonomous vehicles is primarily focused on data vision or using cameras, just like humans, to recognize and understand the world.

Virtually all other companies trying to bring self-driving cars to the road – including Ford, Uber, Waymo and GM Cruise – rely on a series of LIDAR, camera and radar sensors. These companies claim that LIDAR can do things that cameras and radar cannot do, while providing overlapping capabilities to the things the sensors can do. These capabilities, known as redundancies, are extremely important for fully-wheeled vehicles, as they provide an important setback in case of failure.

Photo by Vjeran Pavic / The Verge

Musk's argument that LIDAR is useless with cameras because it replicates the visible light spectrum is "just wrong," says Sam. Abuelsamid, senior analyst at Navigant, a technology consultant. The human eye will respond to wavelengths ranging from 380 to 740 nanometers, while LIDAR can respond to the highest range of 905 to 1500 nanometers. And new types of LIDAR, like the continuous beam FM sensor from Blackmore, can instantly measure speed while reducing computational latency, Abuelsamid notes. Layer sensors with different properties, instead of relying only on a vision-based system, are "ultimately a safer and more robust solution," he said.

The Tesla CEO has been very pronounced and critical of LIDAR's use for autonomy, going as far as calling it "lame" on Monday's event. "In cars, it's freaking stupid," Musk said. "It is expensive and unnecessary. And as [Tesla AI director] Andrei [Karpathy] said when solving vision, it is worthless. So you have expensive hardware that is worthless on the car."

LIDAR can be incredibly expensive, and it is a costly effort for most companies. They can do the job of selling self-driving cars to customers virtually impossible thanks to the extra costs. But companies that use LIDAR are already working to reduce these costs. Waymo started producing its own cheaper LIDAR sensors in 2011. At that time, Waymo said it could lower the device price from $ 75,000 for a LIDAR sensor of just $ 7,500 with its own customized version. In March, the company announced that it would start selling its smallest LIDAR to third parties.

Investors who participated in the Tesla Autonomy Day event were not impressed by Musk's anti-LIDAR tirade. "We continue to believe that Lidar is a good complement to cameras, radar and ultrasonic sensors, and that the ultimate Level 5 winners combine all of these sensors using sensor fusion," wrote Cowens Jeffrey Osborne in a note on Tuesday. "Tesla's rejection of the technology as a" stupid is the credit "because of the high additional cost currently being is probably penny-wise, but pound foolish, especially since the Lidar prices continue to fall."

Musk was not as repugnant to simulation programs as he was of LIDAR, but he claimed that even the most sophisticated simulations fail to capture the ultimate "banter" of the real world. "If the simulation completely captured the real world, well, I mean it would prove that we are living in a simulation, I think," he said, and snarled back to his comments on Recode s Code Conference in 2016 "It doesn't. I want."

His comments seemed targeted at Waymo, the company looks mostly like having the lead in the race to get completely driverless cars. Waymo is great on simulation. It often explains how the Carcraft system has driven 7 billion miles in simulation, compared to 10 million miles on public roads last October. Musk argues that simulations are insufficient to capture actual driving, which means that Waymo will not be able to obtain all the miles the Tesla cars have run with a simulation. (A Waymo spokesman refused to comment on this article.)

Musk also criticized the use of using HD maps to help control self-driving cars through an environment. Before depositing vehicles in a city or city, Waymo first builds a detailed picture of this area and categorizes "interesting features," such as roadways, fire holes and intersections. Knowing the "permanent characteristics of the road", Waymo sensors can focus on moving objects such as other vehicles and pedestrians.

According to Abuelsamid, the benefits of the HD maps come from being able to accurately locate the car in its environment by triangulating the distance from known objects. HD maps also contain data on the directions, such as the speed limit or which paths you can turn off. This limits the area that needs to be scanned by the car's perception system, and lets these sensors focus on what's most important: pedestrians and other cars on the road.

The biggest shock came when Musk unequivocally stated that Tesla would have "one million" driverless cars on the road by the end of 2020. The plan is for you to operate commercially in a driving network. Asked if these cars would be "Level 5 without geofence", which means they could travel anywhere, under some conditions without a human being behind the wheel, Musk said yes.

Some analysts went away with the impression that the robotaxi idea was underdeveloped, or "half-baked", wrote Cowens Osborne. Some experts doubt that there may ever be a genuine level 5 autonomous vehicle. As it is, the vast majority of AV engineers are trying to perfect autonomous level 4 driving, which requires zero entry from a human driver, but only in a particular geographical area or under special conditions (such as good weather).

"For the foreseeable future (if not always), cars or other mobility devices will have to operate under a number of technical limitations," said Bryan Reimer, a top researcher at MIT's Transportation and Logistics Center. expanding the geographical domain for level 4 vehicles, not the "unlimited level 5 problem."

Reimer said that no engineering group working under a "reasonably skilled regulator" would not develop systems running in non-permissible conditions Safe mobility, like snowstorms or hurricanes, Musk suggested that his cars could work in all scenarios, whatever the danger.

Don't worry about the rest of the industry disagreeing. At a technical conference last year, Waymo, CEO John Krafcik said , that "autonomy will always have limitations", which was an outrageous behavior from the leader of the leading driverless car company. about reining in expectations. "We overstated the arrival of autonomous cars," Ford Hackett, Ford, said earlier this month.

It is almost as if the whole industry was trying to make it easier for all the hype and cash that has been dumped in this room in recent years. Everyone except Musk, it is.

"While Tesla will surely continue to introduce better assisted driving skills over the coming years, true autonomy is a completely different game altogether and requires a completely different approach," says Austin Russell, CEO of LIDAR manufacturer Luminar. anymore a backup driver, the hands of the wheel, takes over when things often go wrong. No doubt that it will get [twice, three times] or maybe even [10 times] better over the car's life cycle. But for autonomy, it must be quite literally near perfect … to achieve security that is greater than the average human driver. "

The argument could be made (and this is done in the commentary of former Verge articles about Tesla) that Waymo and other AV companies need to fight it off as Musk to produce the best result for us , the customers. If Waymo is the only game in town, we will probably pay more for rides and be exposed to many forced Google bindings, ads and data sharing about our moves. With Tesla (and Ford and GM) in the mix, Google will be forced to move faster and behave better.

However, another way to look is to bow the rules to what is safe in self-driving cars and what is not, musk puts people at risk right now. Opinion poll believes 71 percent of car drivers around the world can buy a driverless car today, but of course they can't, but they think they can because some car manufacturers are designing and marketing cars in such a way that drivers believe they can renounce the cow The troll Tesla has been criticized for selling a "Full Self Driving" option that confuses what the vehicles are capable of doing.

Musk wants to gain a competitive advantage by referring to Tesla as "fully self-propelled" vehicle, but he burns consumer confusion. And a confused driver can be worse than no driver at all.