Oh yes, the features will definitely mature over time. It's just a question of how much. Will they get mature enough to be real full self-driving or will they mature but fall short of actual full self-driving? Will the features require new hardware or new sensors to achieve real full self-driving? Those are the big questions.
My opinion is that they’ll get to very solid automatic traffic control stops/starts as long as there are no turns involved. The current test feature is already very impressive and is reliable even in extremely complex situations. I could see this happening in summer of 2020.
I think they’ll get auto lane changes working in urban environments too. Manually initiating lane changes in the city is already quite smooth. Maybe late 2020 for this.
I think they’ll get to level 3 highway driving in a single lane. The car is essentially safely capable of this today. Sometimes it’s herky jerky but I’ve never seen unsafe behavior in my car in Southern California and we have plenty of edge cases. Maybe late 2020 for highway level 3.
Then of course they’ll continue to refine navigate on autopilot, summon, smart summon, and autopark to the point that the features approach usefulness. One could argue that NoA is nearing that goalpost today.
Is that full self driving? It’s certainly not what they showed us at the launch of HW 2.0 or at autonomy investor day but it might be enough to get them out of legal hot water with customers who bought the package under more recent verbiage.
I would personally be satisfied with the above suite for my $7,000. I would love to largely ignore my car on the highway and have it react correctly to lights and stop signs.
The glue holding all the magic together though, the merges and turns and yields and anticipating pedestrians and all that...that I’m not so sure about. If they can truly pull that off it will be a huge industry upset but I’ll remain happily surprised if they ever get that off the ground, not angry that I never got it if they fail.
