4 ms·

> https://mobile.twitter.com/taylorogan/status/1469404579439824899 https://mobile.twitter.com/taylorogan/status/146940457943982... > He wants to make people be

> https://mobile.twitter.com/taylorogan/status/1469404579439824899 https://mobile.twitter.com/taylorogan/status/146940457943982... > He wants to make people believe in the illusion that Tesla are in some unsurmountable lead when it comes to autonomy but they are actually playing catchup and are dead last because of the approach they took. I'm sorry, but yeah, the results in that video are worse than what we had 14 years ago after the DARPA Urban Challenge. I'm actually astounded how bad it is. An insurmountable lead this is not, and points to how much of an advantage LIDAR is. Also, while we all have our robot failure blooper reels, this is a system that's live and in production, and Tesla should be ashamed to release this even under the "Beta" label.

mcot2 5y ago
Correct me if I am wrong but the approach used by most in the DARPA challenge was lidar using pre-mapped routes and heavy use of localization. If the route changed significantly I think those cars would have trouble. I belive Teslas approach might currently feel behind but has much more “headroom”. As Elon states the team have gone down many paths that lead to a local maximum which is short of the actual end goal. They firmly belive that an end-to-end system has to be based on vision and has to be able to drive in any situation the very first time without localizing itself to a preset path.
ximeng 5y ago
https://m.youtube.com/watch?v=0wJAANgG-Vg https://m.youtube.com/watch?v=0wJAANgG-Vg You mean this? It looks much less capable and a much more constrained environment than Tesla deals with.
- ModernMech 5y ago
  MIT was in a collision and probably should have been disqualified, so not the best example. Look up Boss from CMU. Moreover, what I said was after the DUC. The results you see at the DUC were the culmination of a little over a year’s worth of work. They were at best prototypes constructed under great pressure in a competition setting. A lot of what you see there was held together by duct tape and 80/20, and the code was not something you’d want to bet your life on. But things did improve rapidly soon after, as results were disseminated, the community began to understand what worked and what didn’t in designing these systems.

mcot2 5y ago

Correct me if I am wrong but the approach used by most in the DARPA challenge was lidar using pre-mapped routes and heavy use of localization. If the route changed significantly I think those cars would have trouble. I belive Teslas approach might currently feel behind but has much more “headroom”. As Elon states the team have gone down many paths that lead to a local maximum which is short of the actual end goal. They firmly belive that an end-to-end system has to be based on vision and has to be able to drive in any situation the very first time without localizing itself to a preset path.

ximeng 5y ago

https://m.youtube.com/watch?v=0wJAANgG-Vg https://m.youtube.com/watch?v=0wJAANgG-Vg You mean this? It looks much less capable and a much more constrained environment than Tesla deals with.

ModernMech 5y ago
MIT was in a collision and probably should have been disqualified, so not the best example. Look up Boss from CMU. Moreover, what I said was after the DUC. The results you see at the DUC were the culmination of a little over a year’s worth of work. They were at best prototypes constructed under great pressure in a competition setting. A lot of what you see there was held together by duct tape and 80/20, and the code was not something you’d want to bet your life on. But things did improve rapidly soon after, as results were disseminated, the community began to understand what worked and what didn’t in designing these systems.