Hacker Newsnew | past | comments | ask | show | jobs | submitlogin

It's not just the software. They are trying to make due without the kind of sensors that are required to properly sense the immediate environment (LIDAR)


How expensive is LIDAR even? My xiaomi vacuum has LIDAR.


They were considerably more expensive the first year Tesla FSD was due to be ready by the end of the year. After that I guess it was deemed easier to do magic with some low quality cameras than get certain people to admit they're wrong and change course.


Current ones still 10-20k, some advances and mass production will lower that further. And there are very different LIDARs (frequency, resolution, range!!), the ones in Smartphones and other small devices work very different.. maybe as different as the ultrasonic sensors in your car vs medical ultrasonic (beware, bad comparison actually).


> My xiaomi vacuum has LIDAR.

I don't. Yet, somehow I operate vehicles with little to no "phantom braking or unintended acceleration."


Teslas don't have these things called "eyes" that have a few millenia of "neural network training"


More importantly, Teslas are lacking a brain.


Indeed. The difference isn't the type of sensor. This oft repeated argument that Tesla's problems are down to a lack of LIDAR doesn't add up.


Hard disagree. Unless they manage to get a computer as good as a brain, they're going to need all the sensor help they can get.


Is there evidence that any amount of "sensor help" can somehow overcome the lack of a limited "brain"?


I'm guessing your vacuum isn't doing 140 miles an hour with four humans in it with an expectation of life-conseving flawless performance for the next 10-15 years. And definitely not in the rain, at night, on rough and uncertain road.


Can your vacuum see 300 meters away and plan how to clean the east wing of your estate?


I live in an apartment so I've yet to really see the reach of this thing's lidar. It maps the room almost immediately though, then cleans by first lapping the perimeter of an area then does laps to cover the inside of the perimeter like someone mowing grass. The map is cached from the last full cleaning, so if I wanted it to do some spot cleaning I can define an arbitrary rectangle and specify how many passes I want over that spot (one is always enough, the second pass never picks up anything else). I can also control it manually but the controls have some latency from the app. I love this thing, it fills its chamber entirely with cat hair that I wouldn't even see until sending it out. It also tracks maintenance and tells me when sensors or the filter needs cleaning, or if parts like brushes need replacing (haven't needed one yet).


Mine most likely can’t see 300 meters, but it definitely knows where it is and plans it route, optimizing over time. It’s kinda neat to pick it up in the middle of vacuuming, move it to a different room, then set it back down and watch it go back to where it left off.


Does it drive you to work too?


This may surprise, but LIDAR is a very recent innovation and virtually zero percent of cars (particularly historically) are LIDAR equipped. Binocular vision has a pretty impressive track record.


Binocular vision as in human vision? Powered by the human brain? That's vastly different than binocular vision powered by Tesla's rudimentary "AI"


Also our binocular vision is movable unlike cars.

Which is how we determine depth when there is occlusion ie. we move our head.


And cars have multiple cameras simultaneously in 360° overlapping coverage. Plus you have motion parallax.


Neither of which completely solves the problem of how to determine the depth of objects in the scene.

We move our head to ascertain depth. LiDAR uses laser pulses.


And yet their self-driving still sucks compared to what a person can do.


Human eye has the dynamic range of 24 F-stops, and best cameras have like 14. That's a difference of several million. Also humans have hearing, and really advanced CPU.

Tesla can't tell apart a deadly bollard from a large plastic bag flying in the wind.


I didn’t think Tesla used binocular (stereoscope) cameras.

These comments seem to confirm: https://teslamotorsclub.com/tmc/threads/does-tesla-have-bino...


They decided to forgo LIDAR specifically though, it's not that it hasn't been out very long and they haven't adopted it yet. Other car makers have been using it for a while, just as long as Tesla could have if they truly want to be cutting edge.


It might also come as a surprise that the the lack of a LiDAR, and famously disabling the radar directly relates to Tesla’s poor self driving performance compared to the actual leaders in the area like Waymo and Cruise.


Not with electrical brains, unfortunately.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: