Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Cruise's 3rd generation self-driving car

medium.com

Post reply on HN
Print view View original

191–200 of 236 posts

Re: Cruise's 3rd generation self-driving car

#191

Post by bobsil1 »

bitL wrote: post #187
bobsil1 wrote: post #186

Earlier quoted context omitted.

The edge cases are unclear: construction zone, heavy equipment, school kids, another car poised to enter the road.

They are probabilistically modeled already. You keep tracking other objects and compute various possibilities for their movement and depending on the past and present behavior pick the most probable one. If car encounters lower-than-threshold probability, it can always slow down or stop, like what non-reckless humans would do. Even end-to-end deep learning can recognize construction zones and learn to drive around tr…

The edge cases are endless (wheelchair woman chasing a chicken), and without a semantic world model the car will have no idea what to do. Low-level object avoidance architecture isn't enough.

Re: Cruise's 3rd generation self-driving car

#192

Post by cmrdporcupine »

revelation wrote: post #101
1_2__4 wrote: post #38

Earlier quoted context omitted.

Maybe stop believing what companies say and start looking for third-party evidence that they're telling the truth

That gets boring fast, if you go by the evidence the only player in the self-driving space is Google. Ironically they are the most restrained in public statements while there is a long list of pretenders announcing they'll ship in 2020 and already discussing far-fetched (entirely) philosophical scenarios, like the whole abortion of an discussion we had on "should the car hit pedestrians or save its driver". The whole…

Thank you, this is exactly spot on.

My personal skepticism on this subject has made me very unpopular at parties (well I don't actually go to parties, but you know) where lots of techy types go. Also earned me lots of downvotes here.

Re: Cruise's 3rd generation self-driving car

#193

Post by bobsil1 »

skadamat wrote: post #178
madmax108 wrote: post #128

Earlier quoted context omitted.

I've been speaking to friends working at different companies across the self-driving industry, and almost all of them have their fair share of skepticism of L4 or L5, including those who are actively researching in the field. The main reason for this is that current self-driving tech has plateaued with most companies going with the redundancy based approach (let's have 3 LIDARs instead of 1, and make that 4 IR sensor…

I recently read about Luminar ( https://www.luminartech.com/ ) and their focus on making affordable, accurate LIDAR tech. Wonder if there are other components like this that need to be optimized and combined to get to L4 and L5

Sensors help, but the real bottleneck for edge cases is artificial general intelligence.

Re: Cruise's 3rd generation self-driving car

#194

Post by bobsil1 »

revelation wrote: post #101
1_2__4 wrote: post #38

Earlier quoted context omitted.

Maybe stop believing what companies say and start looking for third-party evidence that they're telling the truth

That gets boring fast, if you go by the evidence the only player in the self-driving space is Google. Ironically they are the most restrained in public statements while there is a long list of pretenders announcing they'll ship in 2020 and already discussing far-fetched (entirely) philosophical scenarios, like the whole abortion of an discussion we had on "should the car hit pedestrians or save its driver". The whole…

They're bulk uploads over home WiFi, and Tesla hired plenty of ML names like Andrej Karpathy. Having said that, AGI is the bottleneck for full self-driving.

Re: Cruise's 3rd generation self-driving car

#195

Post by bobsil1 »

RivieraKid wrote: post #161
mabbo wrote: post #21

Earlier quoted context omitted.

They've got some damn impressive videos on YouTube. I think they're either neck and neck with, or possibly ahead of, Tesla.

Tesla is the leader leader in terms of released driver-assist technology. But they're years behind in the full-self-driving game. They're no way near where Waymo is now (and most likely Cruise, Drive.ai, maybe others). Their in-house developed Autopilot v2 is still worse than the third-party (MobileEye) Autopilot v1 they got rid of last year.

Waymo is more server-side, requires hi-res maps. Tesla's approach should be more adaptable long-term.

Their AP 2 is actually In-House AP 1.0, of course it does less than MobileEye v 15.0 or whatever.

Re: Cruise's 3rd generation self-driving car

#196

Post by bobsil1 »

inverse_pi wrote: post #152
aerovistae wrote: post #63

While I'm almost always bullish on new tech, self-driving cars are one where I frankly have a pretty hard time buying into the hype. I think when they really start to hit the streets, some issues are going to show up. Here's what I mean. Imagine this situation: You're pulling into your driveway in your self-driving car, and one of your kids left a lawn gnome there, blocking it. The car won't go forward because there'…

You're thinking about the tail end of the distribution. If self driving cars can solve 90% of the mass of the distribution, i.e. driving on highways and urban streets, it would be a significant improvement in terms of efficiency. Of course there will always be morale questions like the one you proposed. But quite frankly, the return on investment of solving the tail end scenarios like that is really small.

You can't ship without restricting where the car goes, punting edge cases to remote drivers, or actually solving the tail end with AGI. Actual full self-driving is far more satisfying than "you can go from here to the airport" or "you can drive in downtown SF."

Re: Cruise's 3rd generation self-driving car

#197

Post by senatorobama »

adrianmacneil wrote: post #176
cruiseemployee wrote: post #48

deleted

Cruise employee here. I'm sorry to hear you feel that way. In my experience, Cruise is one of the most exciting companies to be working for today, and Kyle is easily the most intelligent, hard working, and reasonable CEO I have worked for. For sure, there is still a lot of work to be done before we can launch a full L4/L5 vehicle, but I would say with high confidence our platform is ahead of most other players in the…

This sounds a bit like damage control to me, but

> Kyle is easily the most intelligent, hard working, and reasonable CEO I have worked for.

Couldn't you be in a different management hierarchy to this purported employee?

Re: Cruise's 3rd generation self-driving car

#198

Post by bitL »

felippee wrote: post #188
bitL wrote: post #181

Earlier quoted context omitted.

Waymo had <100 situations where they required human involvement in tens of million miles they drove so far. They are probably already safer than human drivers and would likely cause fewer accidents. The issue there is a legal one - who is going to be responsible for a crash? The company that produced the driving software/hardware, or the owner of the car? Or even the passengers in case there will be city-wide car poo…

No. They had plenty more. Approximately 100 cases over 600k miles last year (their own data from California) which would have lead to a dangerous situation, and 9 cases which by their own admission would have resulted with a collision. Now it is hard to compare crashes of different magnitude, people often are engaged in fender benders, but only on average after 100 million miles driven there is a fatal accident, who…

Alright, I might have misremembered as last time I've read about it I was surprised how low those numbers were. Maybe I need to look at it again.

Re: Cruise's 3rd generation self-driving car

#199

Post by bitL »

bobsil1 wrote: post #191
bitL wrote: post #187

Earlier quoted context omitted.

They are probabilistically modeled already. You keep tracking other objects and compute various possibilities for their movement and depending on the past and present behavior pick the most probable one. If car encounters lower-than-threshold probability, it can always slow down or stop, like what non-reckless humans would do. Even end-to-end deep learning can recognize construction zones and learn to drive around tr…

The edge cases are endless (wheelchair woman chasing a chicken), and without a semantic world model the car will have no idea what to do. Low-level object avoidance architecture isn't enough.

Actually that case would be detected as two objects on a collision course and car would stop/slow down. Mind that sensors scan surroundings with high frequency and some of them can see "beyond the walls" like radar. Ultrasonic sensors can also detect the state of road surface (dry, wet, icy etc.).

Re: Cruise's 3rd generation self-driving car

#200

Post by bobsil1 »

bitL wrote: post #199
bobsil1 wrote: post #191

Earlier quoted context omitted.

The edge cases are endless (wheelchair woman chasing a chicken), and without a semantic world model the car will have no idea what to do. Low-level object avoidance architecture isn't enough.

Actually that case would be detected as two objects on a collision course and car would stop/slow down. Mind that sensors scan surroundings with high frequency and some of them can see "beyond the walls" like radar. Ultrasonic sensors can also detect the state of road surface (dry, wet, icy etc.).

Yes, but when can the car proceed, and in which direction? Helps to know what a wheelchair, a woman and a chicken are and how they behave. Helps to be able to speak with them.

Superhuman sensors will be great, but you really can't dodge AGI for true full self-driving.

Post reply on HN