Tesla Confirms That the Full Self-Driving Beta Will Never Be Full Self-Driving
jalopnik.com
jalopnik.com
My gut reaction is that we will see more along the lines of people watching movies/drunk autopiloting/edge case accidents that cause federal investigations/class action lawsuits/other things that would really put a dent in Tesla as a company and possibly stop them from accomplishing what I see as an amazing task.(start a new car company this day in age, seriously advance driver assist despite calling it more than it really is)
That said, Jalopnik has always been a sorta tabloid car mag that now seems to be largely staffed by people who resent cars to some degree. I will make an exception for the author of this article as Jason Torchinsky has always been a pretty good writer and is one of the few left at the site worth reading. Hes always been quick to pile on the negativity towards Tesla even when he says he is defending them or otherwise praising their products. I dont really have a problem with that, per se, but he has a clear anti Tesla bias where other manufacturers can seem to get more of a pass from him.
My personal opinion of Tesla is a mixed bag, specifically because they seem to have a branding/marketing problem despite my appreciation for them legitimately being the company that is trying to push car engineering forward and past ICE. I totally respect that.
My biggest issue is that it seems like regulations are being wielded as weapons against those that aren't in the good graces of the government for whatever reason. Companies have gotten crucified for false advertising claims and Tesla seems untouchable. People go to jail over securities fraud and Musk gets a slap on the wrist and is mouthing off on TV the next month. He's supposed to have a twitter sitter right now per his securities fraud agreement, would any reasonable lawyer have allowed him to tweet about gamestop or dogecoin?
Of course, I'm also skeptical of the tractability of the problem of actual self driving in the full range of conditions humans handle. It might end up one of those unfortunate problems where many different development paths get to 97% but none of them get to 100%.
My hunch is they do this to be more precise in terms of what the software actually controls.
Having previously worked in safety critical software roles, there were times when the team would get into disingenuous word smithing to avoid safety requirements. E.g., “the software isn’t ‘autonomous’ because it requires a human to make a destination decision at the onset”. In this case, using the word “dynamic” removes any ambiguity that the software is making continuous, time sensitive decisions after the human. Just calling it “driving software” probably isn’t prescriptive enough
I have never seen a Tesla advert where 'the driver gets in the car outside his home on City-A, presses a few buttons (a la Star Trek), falls asleep, the car starts driving to his/her office on City-B and honks to wake him/her up when it self-parks'
So.. features? (not picking a fight with you)
For cars to be fully self-driving they need to talk to one-another. Not just sensors, sensing, but also have the (regularoty mandated) transponders so that each vehicle notifies other vehicles of its position.
I remember when I was learning programming (many many years ago) the exercise to write the code where an object of defined dimensions AxB needs to move from Start to End in a closed circuit, with scatered obstacles inbetween. For me this is how a car needs to do. Tell other cars 'hey these are my dimensions, my current location, my speed, my direction'. With the current computing power it would be easy for cars to avoid one-another (to be proven ofc).
Though I wouldn't mind having such a transponder on my bicycle though if I had the option, or a "personal anonymised transponder". The problem with that implementation is that it's a nightmare to secure it. Blockchain may be used for that?
What comes to mind is the 'Fifth Element' movie, scene when Bruce Willis tampers with the electronics of this cab when police is chasing him, and he's burning the little 'points' he has left.
Perhaps an early implementation could be limited in specific parts (specific highways, when you are 'woken up' before the entry & exit' and the car takes over for the long stretch. A combination of transponders, sensors, magnetic beacons/'drivers' on the asphalt, GPS, radio antennas (kinda like Android/Google maps use for location).
I tend to NOT buy something until proven to work. Musk is a media/marketing freak. He will tell whatever he has to, to serve his business interest. It's down to us to jump the bandwagon, OR wait calmly until the 'thing' is actually produced, tested, proven.
Even assuming you have developed a standard protocol that can be used for cars to communicate with each other, how do you make sure they coordinate? How do you protect against malicious actors?
The programming analogy only works up to a point, the general problem would be coordinating multiple independent agents that you don't control, without knowing the code they're running.
Either put cars on rails like maglev that routes them around foot traffic (still not perfect - people do die by trains, but more controllable - usually it's their dumb fault or suicide), OR hover cars that fly about 15 feet off the ground - high enough to dodge 99% of obstacles/bikes/pedestrians/etc. They could even "stack" vehicles then if there's congestion one group @ 15 feet the next at 45 feet.
Sure there's a lot of tech required for that, but it seems easier to pilot drones by ai than it does cars on streets with so many variables unknown.
So whatever people think of Tesla being too slow, Elon has the biggest incentive to make FSD working as soon as possible.
Personally he said that China wants him to enable it end of this year, and we know that China can arrange things even if it compromises safety (let's hope it doesn't increase the number of accidents).
Also a trip with Tesla FSD Beta is almost 2 times faster than Waymo right now, because it takes left turns as well, and drives aggressively. I expect Tesla to turn left turns off for FSD first.
Tesla regulatory filings differ from advertising.
Jalopnik is notorious for bashing tesla and shilling for their advertisers.