Oy... I never know where to fall on George Hotz. On the one hand, he's done some cool stuff. On the other, he was shilling a device that he more or less insinuated was as good as Autopilot using just your phone which wasn't true for a myriad of reasons I won't get into but beyond that he seems to (at least at the time of this interview) have a core misunderstanding about how Tesla is approaching the problem.
His whole beef is that yeah you can make the car but it's a $250,000 Waymo car outfitted with LADAR and a bunch of expensive sensors. He calls the Waymo cars a press demo and yes, that's exactly what they are. They are press cars and engineering vehicles. Historically vehicles like that cost a ton of money just by virtue of their intent. People used to do this with the Model S. They'd say well Tesla is selling them for $70,000 but they cost $100,000 each to make, not taking into account economies of scale. The Waymo cars aren't built with profitably or economies of scale in mind. It was never their intent as far as I know.
But beyond that you look at Tesla and how they're approaching the problem and it's not at all what he describes. Tesla is building a map of it's surroundings both in-car in real time and utilizing a map and white/ black lists of objects and signatures from a network of over a million cars and growing faster by the day. At autonomy day Tesla showed what the cameras are able to do as far as mapping surroundings and between them and radar they can come up with a 3D world. That kind of lends some credibility to Elon's claim that LADAR isn't necessary.
Will Tesla actually achieve any of this any time soon? Who knows. On the one hand they said in 2016 the cars would be driving themselves a year later. Not only has that not materialized but core features are still missing and they've gone through two hardware revisions since that original release and we don't appear to be any closer. There is no functional difference between the full self driving cars and the cars with enhanced auto-pilot and even the enhanced auto-pilot cars haven't received all of their promised features yet. Namely auto-parking and smart summon that doesn't suck.
On the other hand, I hop on the expressway with a destination entered on my navigation and watch my car merge into traffic for me, observe it as it overtakes slower traffic in the passing lane and then gets back into the slower lane, moves over in the HOV lane when appropriate, switches lanes to follow my route, places me on another expressway and then leaves the expressway at my exit all without my hands on the wheel or any intervention from me and that is truly uncanny. The same hardware that two years ago could barely keep between the lane lines on the expressway is now effectively fully autonomous for highway driving and the only difference between now and then is software. My car is number 431 off the assembly line and is running the original version 2.5 hardware. The level of improvement over the past two years is just unbelievable. The car is able to handle 99.9% of highway situations and in most cases doesn't handle them like you'd expect an autonomous car to do. Lane changes are assertive.
And you look at the app while you're using Smart Summon and your phone shows a digital representation of what the car sees around itself using it's sonar sensors and you watch is navigate based on that info and it's almost silly to see. It's not great but it's so close to great that you know in a few software updates it will be.
So at least as far as Tesla goes, he's off base. And also as far as Tesla goes, what they've done and how quickly they've come since splitting with Mobileeye (AP1) in 2016 is incredible right up until you compare it to Elon's claims. If Elon wouldn't have promised the moon it would be a different story and I wish he'd stop saying this year is the year they go full self-driving because he's said it for 4 years now and has really lost all credibility with respect to timelines.
