Earlier in August, Tesla CEO, Elon Musk got Tesla fans excited about a complete rewrite coming of the Autopilot feature. Using Twitter to set a date, he said a limited public release would be out within 6 to 10 weeks. So, what exactly will a rewrite of Autopilot entail? According to Musk, these aren’t just tweaks, it is a quantum leap:
The FSD improvement will come as a quantum leap, because it’s a fundamental architectural rewrite, not an incremental tweak. I drive the bleeding edge alpha build in my car personally. Almost at zero interventions between home & work. Limited public release in 6 to 10 weeks.
However, this week, the Tesla boss got quizzed about the progress of the update, to which he responded that Tesla owners could get their hands on it mid-December…if there are no major setbacks:
Releasing private beta in 2 to 4 weeks, public beta (early access owners who opt in) 4 to 6 weeks after that, then all US Tesla owners mid December. Above schedule is contingent upon not encountering major unexpected setbacks.
Considering that Autopilot is going through a total overhaul, it will no doubt require thorough testing so having to wait a little bit longer is all for good measure. What we do know is the upcoming rewrite will bring in a lot of new functionality all at once, and pave the way for future improvements to the company’s labeling software.
It’s poised to eliminate all previous ‘mistakes’ made with its driver-assist system in the past. It is reported that the ‘new’ Autopilot will combine the video feeds from every camera in the car to create a surrounding 360-degree 3D layout.
This will also need a hardware upgrade. So, because Hardware 1.0 and Hardware 2.0 don’t have the processing capabilities for that much data, we’ll see Tesla’s Hardware 3.0 being instrumental in the Autopilot rewrite. Tesla is reportedly starting Hardware 3.0 retrofits for Model S and Model X that are equipped with the company’s Hardware 2.0 computer.
Musk has always had a vision for autonomous self-driving vehicles. While the concept may be simple, it takes a lot of brain to engineer a real-life version. Basically, you need to outfit a car with cameras that can track all the objects around it and then have the car react to it it’s about to collide with something nearby. And the car’s computer will be taught the rules of the road. However, following a list of rules of the road isn’t enough to drive as well as human do (in some cases). Take for example the simple act of making eye contact with another driver to indicate they go first, or just make judgement calls, which is not easy to hard code into a computer. But with AI, self-learning is a method where systems could teach themselves how to respond and how to drive after analyzing simulated data and driver behavior.
Getting a computer to drive a car is easy. Getting a computer to drive a car on roads shared with thousands of human drivers is something else. But if Musk could invent a rocket returning safely back to Earth in one piece, then his vision is not far-fetched, especially with a team of 300 Jedi engineers working on the Tesla Autopilot. And Musk has commented before that one day, the cars won't even need steering wheels.
It may seem deceptively smooth and easy for Tesla owners to use the Autopilot’s advanced driver-assist and self-driving features, but it’s been an incomprehensible amount of work to get this far. It just takes a bit of imagination to realize what effort is going into rewriting the entire architecture for an even more advanced self-driving function than we've ever seen before.
