Ahh, "Soonish" - the most applicable word to apply to FSD ever. It simultaneously sounds like it means something while having no content at all.
I was too when I got my car in 2016 when FSD was first sold. Still waiting for something that looks anything like what was advertised. I still can't believe they are fighting through city streets autosteer instead of actually doing a useful L3 highway system.
To an extent that is an argument to continually expand the beta test pool (or even rotate it), since as you note the testers gradually learn the cars idiosyncrasies and mitigate them in advance, thus biasing the data. However, with a large enough pool you will get a cross-section of people who do let the car do crazy stuff. And in fact yes, the larger data sets and important for training .. to get really good results you need huge datasets, not only for training but also for testing.
How do you know Tesla is not doing things like this? You keep asserting Tesla is being irresponsible or unethical, but you have provided no real data to back that up. One very clear datum would be the car acting dangerously and causing major accidents, and that hasn't happened. So what is the basis for your claims?
What??? Have you used FSDb?
I’m no NHTSA specialist, BUT…I’m guessing attempting to turn directly into the path OF AN ONCOMING TRAIN qualifies as “acting dangerously”.
You're right .. I saw a human do that only last week and get killed. CLEARLY we should stop ALL HUMAN DRIVERS AT ONCE. I hope you enjoy riding a horse to work. Oh ... wait .. horses can bolt! ... Maybe a bicycle?
nautil.us
Relevant
There are open questions of how many problems "more data" really do solve in this spaceDeep Learning Is Hitting a Wall
What would it take for artificial intelligence to make real progress?
Are you really trying to argue about the word soonish?? Haha. I did put in parenthesis "now or years away" that quantifies (gives content to) my usage
.
Nope. You misunderstood.
). That means better than the worst and worst than the better humans .No, but I am going to joke about it given how applicable to Tesla's autonomous driving it is.
Conveniently, after he had taken literal hundreds of millions of dollars from consumers for it while offering no refunds for his mistake. What a great guy for "admitting" his "mistakes"!
But they're generating so much disengagement data that there's no way they have the staff to analyze it. Even if they were able to analyze every disengagement, how would that help?
I've read that Dojo is in production. They expected to have it online by this year, as it was announced mid-2021. With the current massive global supply delays, especially with silicon chips, I'd expect the project to be delayed. There are major car manufacturers that are stopping production of vehicles because they can't get processors in enough quantity.
Speed limits are the one law FSD is specifically programmed to ignore (now that the NHTSA made it obey stop signs).
To be fair, since there is so much harping on Telsa and FSD, it's important to note that all cars with cruise control, adaptive cruise control, and ADAS systems can exceed speed limits at the drivers request. Even GM's new Super Cruise, and Hyundai's new Ioniq 5 Assisted Driving can go over the speed limit. I don't believe there is any regulation that stops car companies from letting their ADAS systems exceed speed limits.
Nope. You misunderstood.
The goal has been stated as "better than the average human driver by the end of the year [2022]" (this time - the goal has been a moving target
Most people think they are above average drivers, and that may be the case most of the time. Those slight moments where they are tired, texting, talking, makeuping, eating, drinking, yelling (at children), grabbing (for something in the floorboard), daydreaming, spilling coffeeing, radioing, and speeding (#1 cause of accidents/deaths) are when FSD will ALWAYS be better.
I think people are still conflating robotaxi (L5) with "better than the average human...". I'm optimistic but still think we're 2+ years from full robotaxi with Tesla vehicles. Better than the average human by 2022... I'm cautiously optimistic
That's because there are no regulations at all, and because L2 systems are not autonomous at all. Just like you said, "at the driver's request. Autonomous systems have no drivers.
