LNG: TESLARATI

Tesla has released a video demonstration of its Full Self-Driving suite in action. The short video clip follows a Model 3 as it autonomously drove from one destination to another, following stop signs, recognizing traffic lights, and driving on city streets in the process.

Tesla’s Full Self-Driving strategy is centered around custom hardware that the company has developed and a large neural network that continues to get more proficient as it gains more real-world driving data. The recognition capabilities of Tesla’s neural network was on display in the recently shared FSD demo video, as the Model 3 could be seen reacting appropriately once it encountered objects like stop signs and traffic lights.

The ride chronicled in the recently shared demonstration involved zero manual interventions from the passenger on the driver’s seat. Nevertheless, it should still be noted that operators of Tesla’s autonomous vehicles will still need to be observant of the road even when Full Self-Driving becomes “feature complete” by the end of this year, as mentioned by Elon Musk during an appearance at ARK Invest’s For Your Innovation podcast.

“I think we will be feature complete — full self-driving — this year. I would say I am certain of that. That is not a question mark. However, people sometimes will extrapolate that to mean now it works with 100 percent certainty, requires no observation, perfectly. This is not the case,” he said.

Tesla conducted test drives of its Full Self-Driving features after the presentations of Musk, VP of Hardware Engineering Pete Bannon, Sr. Director of AI Andrej Karpathy, and VP of Engineering Stuart Bowers, though the company reportedly did not allow videos to be taken during the test rides themselves. Fortunately, Axosoft founder Hamid Shojaee described his FSD experience in a series of tweets. Shojaee noted that the demo ride of the fully autonomous drive was really impressive. “Parking lot to city streets, to freeway, to city streets back to parking lot, ~10 mile drive, 100% autonomous. Tesla is light years ahead,” he wrote.

Tesla’s Full Self-Driving suite is a key component of Elon Musk’s Master Plan, which calls for the deployment of the Tesla Network’s autonomous Robotaxis that commuters could summon using an app. This will allow Tesla to become a key player in the ride-sharing market, which is currently dominated by Uber and Lyft. Musk noted that Tesla’s Robotaxis will cost around $0.18 per mile, undercutting Uber and Lyft’s $2-$3 cost per mile. Considering the lifespan of Tesla’s battery packs and drivetrain, the Musk expects each Robotaxi to make a gross profit of $0.65 per mile assuming that 50% were empty miles, and an annual mileage of 90,000 miles, resulting in one vehicle earning about $30,000 per year.

Watch Tesla’s video demonstration of its Full Self-Driving suite in action in the video below.

Tesla stock (NASDAQ:TSLA) dropped on Tuesday morning as Elon Musk’s vision for the Tesla Network’s Robotaxi service was met by skepticism from Wall Street. Musk was quite confident in the recently-held Autonomy Day event, stating that Tesla could start deploying its autonomous ride-hailing vehicles by 2020. Wall Street did not seem to be impressed,

Cowen’s analyst Jeffrey Osborne expressed his reservations on Elon Musk’s ideas in a note on Tuesday, describing Tesla’s plans for its Robotaxi service as “half-baked.” The analyst also expressed his skepticism of Tesla’s capability to compete with established companies such as NVIDIA for hardware and Google for software. “We see a significant amount of technology and execution risk in the shift in strategy from competing in just electrification to Tesla also beating Nvidia in hardware, Google in software, and building a better ride-hailing service than current ride-hailing leaders. The Tesla Network Robotaxi plans seemed half-baked, with the company appearing to either not have answers to or not even considered pretty basic question on the pricing, insurance liability, or regulatory and legal requirements,” Osborne wrote.

SunTrust analyst Willian Stein was equally skeptical, though he did state that Tesla’s Robotaxi business could be financially compelling if the electric car maker successfully rolls out the service. “We are concurrently skeptical and hopeful about TSLA’s claims. If TSLA executes to plan, implications across semis/components are meaningful: positive for analog/mixed-signal vendors, mixed for digital vendors, negative for LiDAR vendors. Investors should recognize that, if the company achieves its autonomous driving goals, combining this with its already-achieved EV technology conspire to establish a ride-hailing service that could be quite financially compelling to both the car owner and the rider,” he stated.

Morgan Stanley analyst Adam Jonas admitted that Tesla’s presentation of its full self-driving technology was impressive, though he also noted that Elon Musk’s idea of completely removing the driver (signified by the removal of the steering wheel and pedals on vehicles) could take a very long time. “Tesla’s autonomy investor presentations conveyed impressive technological leadership but, in our view, left big questions around time-to-market and did not change our views on the impediments to removing the human driver in a commercial service at scale. Attention returns to a challenging 1Q. More data to back up safety claims is needed. We still think removing the safety driver will take many years (if not decades) to achieve at high scale,” he wrote.

Not everyone was highly skeptical of Elon Musk’s vision for a Robotaxi service. Loup Ventures managing partner Gene Munster noted that Tesla’s non-LiDAR approach to full self-driving could very well give Tesla a headstart over other companies pursuing autonomous driving technology. “We are more comfortable with Tesla’s camera-based (non-LiDAR) approach to autonomy. If correct, this approach could actually be preferred (safer, more reliable, efficient, better design) and afford Tesla a several-years headstart as other players unwind LiDAR from their solution,” he wrote in a blog post after attending Tesla’s Autonomy Day.

According to Munster, Elon Musk also estimated that Tesla could operate up to 10 million Robotaxis in about 10 years. With an estimated gross profit of $30,000 per car annually (and taking Elon Musk’s estimates at face value), Tesla could be looking at over $300 billion in gross profit in 10 years. That’s a huge opportunity for Tesla, and it could very well make the electric car maker as one of the most valuable companies in the world. Nevertheless, the Loup Ventures partner noted that “given investor optimism on the autonomy theme, we believe this is the right time for Tesla to raise money (debt or equity) to de-risk the story with additional working capital.”

As of writing, Tesla shares are down 0.28% at $262.01 per share.

Disclosure: I have no ownership in shares of TSLA and have no plans to initiate any positions within 72 hours.

Elon Musk did not mince his words when he was asked about LiDAR during Tesla’s recently-held Autonomy Day, an event that featured a deep dive into the company’s full self-driving strategies and initiatives. Ever the bold CEO, Musk noted that LiDAR is unnecessary for vehicles to achieve autonomous driving capabilities.

“LiDAR is a fool’s errand, and anyone who relies on LiDAR is doomed. It’s like having expensive appendices. You’ll see,” Musk said.

Sr. Director of AI Andrej Karpathy elaborated on Elon Musk’s point during his presentation on Tesla’s neural network, stating that Tesla Vision and artificial intelligence are more than enough to perform actions that LiDAR can do. Describing his point, Karpathy lightly joked that the attendees of the event only used their biological neural networks to get to Autonomy Day’s venue without any issues.

“You all used your own neural network in your brains to get here. You didn’t shoot lasers from your eyes to drive,” Karpathy lightly said.

LiDAR is used primarily by most key players in the autonomous driving market. Waymo and GM Cruise, the top two companies pursuing the technology according to research firm Navigant (the firm also lists Tesla as the second-worst company in terms of full self-driving), rely heavily on LiDAR for their autonomous vehicles’ operations. The technology is seeing a lot of support as well, with a Reuters analysis pointing to more than $1 billion in corporate and private investment being plowed into some 50 LiDAR startups in the past three years.

Nevertheless, Tesla is not the only company that is opting out of LiDAR for autonomous driving. Early this month, full self-driving startup Wayve showcased an impressive autonomous driving demonstration using a modified Renault Twizy. Similar to Tesla, Wayve accomplished its feat without using LiDAR. In a statement to Engadget, the self-driving startup noted that its full self-driving technology was made possible by using cameras and teaching AI to drive like a human being.

It should be noted that while Elon Musk has not minced words when it comes to his opinions about LiDAR, he is not completely against the technology. Musk described how SpaceX’s Dragon capsule uses LiDAR because it makes sense in that scenario. But when it comes to everyday driving, Musk was clear on his stance, even predicting that Tesla’s competitors in the full self-driving field will likely abandon LiDAR in due time. “They’re gonna dump LiDAR, mark my words. That’s my prediction. LiDAR in cars is stupid,” Musk said.

Tesla is yet to roll out the full capabilities of its Full Self-Driving suite to its fleet, though Musk has stated that he expects the system to be “feature complete” by the end of the year. Industry leaders such as Waymo are pretty much on the same boat, with most of the companies’ autonomous vehicles still requiring safety drivers as they operate in tightly geofenced areas. Considering Musk’s timeline for the rollout of Tesla’s full self-driving features, and taking into account the potential delays on its rollout, there is a good chance that the LiDAR debate might not be settled for some time to come.

Tesla’s Autonomy Day is starting in a few hours, and the anticipation among the electric car community is reaching a fever pitch. Elon Musk has promoted Autonomy Day over the past week, even lightly hinting that the event will “free investors from the tyranny of having to drive their own car.”

Tesla’s noted in its initial announcement that it will be providing investors a deep dive into its full self-driving initiatives. Several key Tesla executives involved in the development of the company’s full self-driving technology are expected to address investors at the event, including VP of Engineering Stuart Bowers, VP of Hardware Engineering Pete Bannon, and Sr. Director of AI Andrej Karpathy. More importantly, test drives on vehicles that are equipped with yet-to-be-released Autopilot and Full Self-Driving features are also expected to be conducted.

Apart from this information, Tesla has managed to keep the specifics of its upcoming autonomy event secret. With this in mind, here are some developments and features that we expect Tesla to showcase in Autonomy Day.

On April 22, Investor Autonomy Day, Tesla will free investors from the tyranny of having to drive their own car

— Elon Musk (@elonmusk) April 13, 2019

Summon and Navigate on Autopilot Improvements

When Tesla brought back Full Self-Driving as an add-on for its electric cars, the company placed Navigate on Autopilot and Summon, two features that were previously part of Enhanced Autopilot, as part of the FSD suite. Tesla has since improved the features, with Enhanced Summon allowing vehicles to navigate a parking lot to find their owners, and Navigate on Autopilot being capable of changing lanes without any turn stalk confirmation. These features are still being improved by Tesla, and it would not be surprising if the company decides to showcase improved capabilities of Summon and NoA to investors on Autonomy Day.

Traffic Light/Stop Sign response and automatic city streets driving

A look at Tesla’s order page for the Model S, Model 3, and Model X shows that Full Self-Driving has two capabilities marked for release “later this year.” These are “Recognize and respond to traffic lights and stop signs” and “Automatic driving on city streets.” Both these features require Tesla’s neural net to be trained enough to recognize visual input at a level that is quite close to that of a human driver. Tesla appears to be taking steps towards this goal, as indicated by findings from hacker-enthusiast greentheonly, who found what appears to be stop line detection features in 2019.4. Tesla’s recent Autosteer Stop Light Warning, which prompts drivers to take over the vehicle’s controls if a red light is detected, also appears to be a step towards automatic city driving.

Back in the third quarter earnings call, Director of AI Andrej Karpathy noted that while Tesla had already trained larger, more advanced neural networks that work very well, the company is unable to deploy it to the fleet due to the limitations of its hardware. “We are currently at a place where we’ve trained large Neural Networks that work very well, but we are not able to deploy them to the fleet due to computational constraints,” he said. These limitations are expected to be rendered null and void by Tesla’s Hardware 3, which is also expected to be discussed on Autonomy Day.

New mobile app features

At the core of the Tesla experience is its mobile app. It allows customers to view the status of their vehicles, and as shown in videos from members of Tesla’s early access program, it is also a central part of Summon’s more advanced features. A Teslarati reader who is part of the early access program recently sent new screenshots of updates in the mobile app (3.8.3), such as Roadside Assistance as well, which allows owners to call the company for aid in the event of a flat tire, or similar emergencies.

Tesla Network test drives

Tesla’s endgame with its full self-driving feature is the Tesla Network, a ride-hailing service that utilizes a fleet of autonomous vehicles. Back in the Q3 2018 earnings call, Musk noted the Tesla Network will be competing with giants such as Uber and Lyft. The company made its intentions for the Tesla Network even clearer when it released leasing options for the Model 3. In its announcement, Tesla stated that customers cannot purchase their leased vehicles after their lease period is over since the company will be using the electric cars for the Tesla Network. With Tesla expected to showcase its full self-driving features in Autonomy Day, it will not be surprising if the company provides a teaser of what it’s like to experience the Tesla Network firsthand. 

An emphasis on human vigilance

While Autonomy Day will be focused on the company’s full self-driving features, Tesla will likely continue to emphasize that its vehicles and their driver-assist features still require constant attention on the road in the near future. This is something that Elon Musk has teased previously at an appearance in ARK Invest’s For Your Innovation podcast when he described how Tesla’s full self-driving features will be feature-complete. “I think we will be feature complete — full self-driving — this year. I would say I am certain of that. That is not a question mark. However, people sometimes will extrapolate that to mean now it works with 100 percent certainty, requires no observation, perfectly. This is not the case,” Musk said. Thus, for the following years, at least, Tesla will likely focus on Level 4 autonomous solutions, which are capable of operating a vehicle without human input but are equipped with pedals and controls for manual overrides just the same.

Tesla’s Autonomy Day will be livestreamed by the company.