November 3, 2018 by Zachary Shahan
Last week, there were some big Tesla autonomy announcements. First, Tesla has released a feature to navigate the autopilot. Basically, the update allows Tesla vehicles to transition from the ramp to the off-ramp without driver intervention. It depends on the route you have in the Tesla navigation system and the magical Tesla engineers behind this touch screen.
Recently, Elon Musk has told us that you can soon walk like a dog on your Tesla and steer him like a kid's remote-controlled car. These fun and useful enhancements are provided by an update to the Summon feature.
It's cool to explore the possibilities of each of these improvements and be excited, but there is something more dramatic waiting in the shadows between these two announcements. There is something hidden between the parking lot and the highway. Basically, we are a few autonomous ways when Tesla vehicles are completely self-propelled vehicles.
Think about it: your car can navigate alone in a parking lot to come to you. It can also drive off-ramp to the off-ramp on interstate traffic lights (well, if you have taken enough time to do what it feels you are doing at critical times of the journey). What is in the middle of these two driving atmospheres? In some situations – for example, a hotel near the highway and a house or work station near one – almost nothing blocks autonomous door-to-door driving. In more complicated situations, the car has to find its way through the cities, following the usual traffic rules and whatever unusual events happen.
This is also the point where Tesla's self-propelled technology distinguishes the car software battery solar company from companies that are commonly seen as competitors in this field, such as Waymo (aka Google / Alphabet). Tesla uses a fundamentally different system. Waymo's self-propelled system is based on precise mapping and knowledge of traffic rules and frequent traffic events in a relatively small geographical area. Tesla's self-propelled system is based on the cars learning how to drive and react to traffic rules and surprises just as a person would. In other words, it's about driving more like a human than a robot with really good maps.
This gets a bit confusing, and after talking to self-propelled tech experts, I found that even experienced outsiders do not have a good idea of Tesla's behind-the-scenes activities. We know that Shadow Learning is underway – Tesla's autonomous driving systems learn daily from their human drivers – but how much do they learn and what details do they pay attention to? We know that Tesla uses neural networks to speed up and improve the learning process, and that new hardware will greatly improve these neural networks, but we do not have much clarity about the practical phase of these neural networks and the challenges they pose and their shortcomings. or medium term predictions for Tesla autonomy.
Tesla autopilot has traveled 1.48 billion miles (updated estimate): https://t.co/ZOfZn6HlFY pic.twitter.com/7Ze0t7Bft4
– Lex Fridman (@lexfridman) October 9, 2018
Updated data on total deliveries of Tesla vehicles by autopilot hardware: 443,000 total Tesla, 279,000 version 2, 114,000 version 1, 49,000 without AP hardware. @elonmusk https://t.co/v34JLgLIyg pic.twitter.com/J0xukW3nMt
– Lex Fridman (@lexfridman) October 8, 2018
Recent data from Tesla and MIT indicate that the Tesla autopilot has traveled 1.5 billion miles. It has also recently become known that self-propelled Waymo vehicles have traveled 10 million kilometers. It has been a while since I attended a math class, but I remember billion much bigger than his million,
As I said above, these systems are not really the same. They basically use different methods to pursue a self-driving future. However, the Tesla system is that it learns from people how to drive like humans – except safer thanks to better attention span, faster response times, and better vision. Apparently, the system is now good enough to handle parking lots and highways. As these autonomous miles grow exponentially, you can assume that Tesla locks up the few remaining points between the parking lot and the highway.
At the last quarterly teleconference, Elon Musk pointed out that Tesla's hardware for Dwarf autonomous driving is the hardware in "comparable" cars. It prepares for this "small" jump Drive around the parking lot and drive on the highway to drive everywhere, Perhaps people who paid for Full Self Driving at the time of their car purchase will be a bit impatient (the same thing happened when the autopilot was first rolled out), but any irritation about the wait is likely to be erased when Tesla sends the over-the-top -Air update that allows owners to download and use the first versions of Full self-drive, If the same drivers are picked up from work a few years later or their car earns some money by driving people around when it's boring, I think they've forgotten they had to wait a few months or even a year to eat They are the first fruits of their Full Self Driving purchase.
You may recall that Musk stated in July 2016 that Tesla's autopilot system requires at least a billion miles of real driving to exit beta mode. However, it was not clear how many kilometers would need more than 999 million Tesla. "Not enough data is available for less than 1 miles. 1B is a necessary but not necessarily sufficient condition, "Musk tweeted at the time.
Are we already there? My understanding is that the answer is essentially yes or almost.
Apart from the sheer tech aspect and when Tesla feels ready, there are, of course, regulations that require different levels of self-drive, including full self-drive, without a human operator (aka driver) being attentive and capable of doing so Take control when needed. Tech may be ready next year, but when will you do that? Really Are you allowed to send your car to pick up avocados or shuttle passengers for a few euros in the city?
We'll see, and chances are that Elon will break the news on a tweet.