I have been very impressed with the steady incremental real-world improvements to AutoPilot over the last 13 months since I got my Model 3. They have added up to a dramatically more adept system which has become an extremely competent highway driver handling lane changes, merges, being cut off, handling poor visibility, lane closures / traffic cones, even things like scooting slightly over to one side of the lane to give a bit more room when passing a large truck.
One example of behavior which has gotten better over the last couple months is during merges where the right lane would appear to widen, prior versions would steer right to try to re-center and then have to steer left to get back in lane, where now the car understands how to keep its path in these situations, even if there isn’t a square-dashed-line drawn to indicate the right side of the right lane continuing.
I think the majority of the miles driven on my Tesla have been driven with AutoPilot. (Actually that would be a really cool feature if the car could show that number.) It’s become indispensable to me, and it keeps getting marginally better every month.
I’m not saying that it is full autonomy, or even that we will get there on the next couple years. But the cumulative progress over the last 12 months has been incredible, and if it continues anywhere near this rate of advancement over the next 24-36 months, I think AutoPilot will be a truly remarkable achievement — as in, approaching a watershed human achievement akin to spaceflight.
>I think AutoPilot will be a truly remarkable achievement — as in, approaching a watershed human achievement akin to spaceflight.
Just want to point out the irony in this statement. Ok sure, fully autonomous self driving may be a great technological achievement, but it will be an achievement that benefits rent-seeking corporations rather than humanity at large. Sure, the benefits of such a service could allow car-free living at a price cheaper than leasing/owning a car, but this is a benefit that will be reaped by those with means (middle class) while the poor will still be riding the bus or walking. It would also make investment in public transportation even less likely, which would actually result in a net benefit for all.
Full autonomous driving will be a significantly greater accomplishment for humanity than you’re giving it credit for.
I’m not patient either.
Moreover, speech recognition isn't a product you can't launch in a faulty state. You can sell a bad product, make a profit, and iterate without any consequence.
It's not the same with self driving cars, which won't be profitable till you get rid of the driver entirely. And you can't launch a ok product and iterate. It has to be almost perfect when it comes to safety.
Barring geofencing or legislative changes, you have to get it in a perfect state before you roll it out.
When the VC money dies down, and when such big companies are no longer optimistic at having a viable self driving car, the research is going to slow down drastically, till a new wave of funding comes along from some other boom to resuscitate the research.
I think Tesla is doing just this. They are putting self driving features on their cars, putting the onus on the driver to supervise, and using the experiment to gather data. I don't expect their current hardware to ever be able to deliver full autonomous driving, but it will likely provide a useful stepping stone.
No, it merely has to be as good as humans, though it should be mildly better. The massive economic and feature advantages it will bring to the table mean that regular people will paper over problems real quick, just as with, well, current cars. In fact if anything I wonder if a higher rate of fatalities and accidents would actually be psychologically better than "almost perfect", because in the modern environment things that are super rare and low risk tend to get inordinate media attention and political response vs things that happen with a certain frequency and thus people become habituated to. I mean, again cars themselves are the big example of that. In America alone, estimates are around 40,000 died last year, which itself was actually around a 1% drop from the previous few years. Which means hundreds of fewer deaths. But that didn't register remotely like a handful of high profile super unlikely dramatic deaths would.
But the bonuses are the main things vs any negatives, because humans are good at motivated reasoning to deal with risk when it's something they want. And a reliable included always available personal chauffeur that means things like car access for kids without disruption of parents' schedules, continued car access for the elderly (note, one of the most powerful and reliable voting segments of the electorate), elimination of issues due to drinking/tiredness/phone usage, just plain letting people do other productive things or flat out sleep on trips, etc etc etc., is a powerful set of incentives to quickly move past someone else's "bad luck".
Yes you can, tesla and comma.ai doing it right now.
I regularly yell at Google when I ask it to play me a song, oftentimes even when I spell out words letter by letter, painstakingly slowly, and it still fails me.
Speech recognition doesn't work all that well for me.
Anyway, I think nuclear fusion is a much better comparison. There is an inflection point where it just works, and until you reach that it's not really useful to anybody except as a research problem. And based on what I hear from my peers who work on self-driving cars (including at Cruise), it's still decades away. Most of them already quit because they just stopped believing it was ever going to happen.
Well, you catch more flies with honey than with vinegar.
I think we should just "move fast and break things"... make all cars self-drivable via speech recognition. I mean that would would serve to limit car ownership... to those with a death wish.
I live in NY. About 40 miles as the crow flies from NYC. Public transportation is unusable in my town... buses run about once an hour, give or take an hour on weekdays--forget about Sundays. How would you like trying get a sick child to a doctors appointment then going to a pharmacy? Need milk before a snow storm.. good luck. My point is the Vast majority of the USA is not geographically located in metropolitan areas.
As a side: Metro North which I used when I worked in queens... is fricken expensive as hell and slow as molasses.
Oh and one more thing; there are no sidewalks nor lighting on any of the residential streets in our town. When you walk here after dark you better damn well have a flashlight in your hand (most cellphones don't throw enough light far enough), futhermore if its rainy and foggy like it often is in summer or late autumn--dude it's no joke. My main concern when I drive here after dark is to Not hit a pedestrian, they are few and far between but still. My point is; we need car ownership here, will it always be this way? IDK, but that is the current reality and you're not going to change that anytime in the near future without years long infrastructure projects.