After the blog post yesterday and the press conference about Autopilot, some were disappointed that nothing was said about the next generation of the system. The event was focused on Tesla’s new radar processing technology and Elon Musk only answered questions about the update.
Nonetheless, I managed to ask Musk a question to help us have a better idea of the timing for the next generation hardware, often referred to as Autopilot 2.0, by looking at the limits of the current generation.
During his opening statement for the press conference, Musk briefly talked about what was needed for Tesla to implement the new radar processing technology on the current generation of Autopilot. He said that Bosch, its radar supplier, gave them new drivers to access more raw data from the front-facing radar and that Tesla wrote software to process those new inputs.
He added that its was a challenge to make the new software work on the somewhat limited computer power in the current Model S and X. Tesla’s Autopilot 1.0 relies in part on the Media Control Unit (MCU) to relay information from the Autopilot to powertrain controls, but that wouldn’t work for 2.0. The system will need a dedicated computing channel.
Therefore, if we know when Tesla reaches the limit of the current system, it gives us a good idea of when it plans to introduce the next generation hardware considering they will need it to make any further major improvement to Autopilot.
When I asked if Tesla reached the limit of the Model S and X processing power with this new 8.0 software update, Musk admitted to “almost reaching the limit of the hardware”:
“It was a hard software problem. I think in term of macro, major improvements, yes, we are almost reaching the limits, but it is important to emphasise that the fleet learning will continue and the intelligence o f how that fleet learning is applied to the car will continue to improve.”
He added on the difference between the hardware in the car and fleet learning:
“So while we are reaching the limit of the hardware, I think we have not quite yet reached the algorithmic intelligence on the car and of course, anything that’s done on our servers – where we are not computer constraint or space constraint in any way. I think, in fact I’m quite certain that it would continue to improve quite a lot just because the software and the data would improve quite an enormous amount.”
He said that the improvements will be noticeable on the existing hardware “for years to come”. It is good news for current Tesla owners, but it is also clear that Tesla will need to introduce more hardware in its vehicle soon if it wants to develop more Autopilot features.