Tesla’s upcoming updates to its Summon feature are expected to be rolled out within the next few weeks. As noted by Elon Musk in an update on Twitter, the advanced Summon update is currently going through final validation and regulatory approval. Thus, members of the company’s early access program would likely receive the new feature in a few weeks.
Some of the upcoming updates to Summon were teased by Elon Musk last November, when he noted that the improvements would be compatible with all cars produced in the past two years. In a series of posts on Twitter, Musk mentioned some features including the ability to operate a vehicle like a big remote control car. Musk also hinted at capabilities that would allow vehicles to follow their owners “like a pet” as long as they are holding down the Summon button on the Tesla app.
If Elon Musk’s recent tweets are any indication, though, it appears that the company is now at a stage where it is already refining the upcoming Summon updates. While responding to Tesla owner-enthusiast Ryan McCaffrey on Twitter, Musk showed a notable degree of excitement, even dubbing the upcoming features as “trippy.”
Going through final validation & regulatory approval. Probably releases to early access program owners in a few weeks. It’s trippy!
— Elon Musk (@elonmusk) January 10, 2019
On a rather interesting note, Musk has noted that the advanced Summon features would likely not be available in all regions when it gets released, due to “some regulatory pushback.” Musk has not mentioned the regions that are resistant to Tesla’s upcoming advanced Summon features, though this would suggest that some upcoming features, such as the highly-anticipated “remote control mode,” would not be available to all regions.
While arguably still in its basic iteration, Summon remains one of the most notable features of Tesla’s electric cars, particularly as it allows customers to operate their vehicles without anyone on the driver’s seat. Granted, Summon is only capable of moving a vehicle forward or backward for a maximum of 39 feet in a straight line, but it is still invaluable when parking in tight spaces nonetheless. Even former Top Gear host Jeremy Clarkson, who has clashed with Tesla and Elon Musk during the days of the original Roadster, found Summon as a very compelling feature in his review of the Model X P100D during the second season of The Grand Tour.
Over the past few months, Elon Musk has provided teasers as to how the upcoming improvements to Summon might work. According to Musk, advanced Summon would be using the vehicles’ Autopilot cameras to navigate themselves. This is a departure from the system used for Summon’s current iteration, which uses the vehicles’ ultrasonic sensors when maneuvering. Perhaps most importantly, Musk also suggested that in the future, even more advanced versions of Summon would be able to read signs around parking lots to determine if a parking spot is valid or not.
For those unfamiliar, this uses Tesla Autopark/Summon. Slightly smarter version hopefully ready soon. By next year, a Tesla should be able to drive around a parking lot, find an empty spot, read signs to confirm it’s valid & park.
— Elon Musk (@elonmusk) October 31, 2018
For such features to become a reality, though, Tesla’s Neural Network would have to learn how to recognize and read signs. Fortunately, in the company’s third-quarter earnings call, Tesla’s Director of Artificial Intelligence Andrej Karpathy mentioned that the company had already trained large Neural Networks that function very well. Addressing the participants in the earnings call, Karpathy explained that Tesla has so far been unable to deploy these larger Neural Networks due to the constraints of the company’s current hardware — something that would be addressed with the introduction of Hardware 3 later this year.
“It is a common finding in the industry, and that we see this as well, is that as you make networks bigger by adding more neurons, the accuracy of all their predictions increases with the added capacity. So in other words, 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,” Karpathy said.