Slate CEO talks accessories, third party parts, Slate Marketplace (TechCrunch interview)

[KevinRS]

The test of these systems is do they prevent more accidents and deaths than they cause. In most cases they have, with the possible exception of people who have gone full hands free and inattentive in cases where they shouldn't have, likely in violation of how a system is supposed to be used, for example defeating the detection of hands on the wheel or eyes forward.
The Slate is designed to have Automatic emergency braking, Pedestrian Identification, and Forward Collision warning. Those have been listed on the specs page from the beginning. At least part of those will within a few years be required to be standard on new vehicles anyway.
The cameras and sensors required for those should make the version of cruise control where traffic in front of you slowing down gently adjusts your speed down not much more than software.
Lane keeping assist would probably require more sensors.

 

[atx_ev]

i want self driving but in a van type vehicle that I can take the family on road trips. When Im driving around town I dont need self driving as the trips are short. My ideal road trip car is a minivan style car with tesla self driving and EREV.

Until then I want the cheapest EV truck possible.

 

[E90400K]

That's an interesting perspective, I suppose I just see it differently.

The only reason I inferred that you would know the limits of the technology and its intended use environments was because you said "I'm no stranger to automation software development and systems that work with such software. I worked for nearly 30 years in systems integration engineering in a business management roll. Safety of life systems and in the transportation sector for the last 15 years of my career." I assumed from this that you would have a robust understanding of the limits of the technology and its intended use environments.

"Am I supposed to read the entire operator's manual before I drive the car when I have deadlines to get to my office as soon as I can?" When driving a motor vehicle on public roads, you are responsible for being able to safely operate that vehicle and avoid putting others in danger, regardless of if you are in a rush to get somewhere. If you need to read the entire operator's manual before driving the vehicle in order to safely operate it, then so be it.

As far as the technology being "immature" because it has known limitations, that would mean that every single part of a car is “immature”. Anti-lock brakes (ABS) have known limitations and cannot instantaneously stop a moving vehicle. The technology is “immature”. The traction control system (TCS) has known limitations and does not guarantee you to gain traction every time you use it. The technology is “immature”. The Vehicle Stability Control (VSC) has known limitations and does not guarantee the car will follow your intended path of steering. The technology is “immature”. I think you get my point that this line of thinking could apply to everything.

If the choice is to not include any of these systems because they are not “infallible”, or to include them with instructions on how to use them and the ability to disable them if desired, then I would choose the latter. I think we fundamentally disagree on this choice, and I see no benefit to arguing further on what is ultimately a matter of individual preference, so I will simply rest my case here.

At the end of the day, it’s nice that we both appreciate what the Slate truck brings to the table, and I’m sure it’ll serve us both well.

[edit]
Additional responses. To address your ABS, TCS, and VSC. As a group of technologies they all work off of the brake system and engine throttle mapping controlled by a control module (computer). Yes, the group of technologies have limitations due mainly to tire grip and driver behavior. If the driver brakes too late or drives too fast for the conditions, such as snow or rain, the systems can't overcome the laws of physics. That's not technology immaturity. These three technologies via computer control operate the brake system and throttle system and differential operation in ways that are not available to the driver. The driver cannot operate all four wheel brakes independently nor can he modulate the brakes at threshold braking as effectively as the ABS module can because the ABS module has four channels of hydraulic operation. Nor can the driver direct the differential where and when to apply power to a specific wheel. There are no physical controls available to the driver to operate the vehicle in such a manner.

The lane keeping technology and crash avoidance alert systems I specifically addressed are different technologies intended to take over driving responsibility from the operator. Both are intended to address driver (un)alertness. In the case of the BMW, the car changed its direction AGAINST my intentions. It made a mistake because it did not know what my intention was and misinterpreted the conditions its sensors detected. The driving environment the incident occurred in was exactly what you believe the technology is best operated in, a well marked (albeit dirty) well-defined roadway.

The Bronco crash avoidance alert system also made an error based on its interpretation of the data from its sensors (camera and radar). It falsley alerted me of an impending collision because the topography of the road combined with parked cars on the shoulder of the road between the unmarked pavement and woodline the software was not designed to correctly assess the situation. That is technology immaturity.

 

[Bayfire2441]

[edit]
Additional responses. To address your ABS, TCS, and VSC. As a group of technologies they all work off of the brake system and engine throttle mapping controlled by a control module (computer). Yes, the group of technologies have limitations due mainly to tire grip and driver behavior. If the driver brakes too late or drives too fast for the conditions, such as snow or rain, the systems can't overcome the laws of physics. That's not technology immaturity. These three technologies via computer control operate the brake system and throttle system and differential operation in ways that are not available to the driver. The driver cannot operate all four wheel brakes independently nor can he modulate the brakes at threshold braking as effectively as the ABS module can because the ABS module has four channels of hydraulic operation. Nor can the driver direct the differential where and when to apply power to a specific wheel. There are no physical controls available to the driver to operate the vehicle in such a manner.

The lane keeping technology and crash avoidance alert systems I specifically addressed are different technologies intended to take over driving responsibility from the operator. Both are intended to address driver (un)alertness. In the case of the BMW, the car changed its direction AGAINST my intentions. It made a mistake because it did not know what my intention was and misinterpreted the conditions its sensors detected. The driving environment the incident occurred in was exactly what you believe the technology is best operated in, a well marked (albeit dirty) well-defined roadway.

The Bronco crash avoidance alert system also made an error based on its interpretation of the data from its sensors (camera and radar). It falsley alerted me of an impending collision because the topography of the road combined with parked cars on the shoulder of the road between the unmarked pavement and woodline the software was not designed to correctly assess the situation. That is technology immaturity.

I feel like this is a brand issue rather than a technology issue. LKA and ACC should be assist rather than relied upon. Not much to say about the emergency braking you described but that also sounds like thats on Ford.

 

[Dorbiman]

The same argument could be made the other way too. For every time that LKA or AEB has been used and avoided a collision or incident, the human is the immature technology. But if you only consider the times that it didn't work, it paints a different picture.

 

[E90400K]

The same argument could be made the other way too. For every time that LKA or AEB has been used and avoided a collision or incident, the human is the immature technology. But if you only consider the times that it didn't work, it paints a different picture.

Yet, John Connor ultimately won the war...

 

[Daemoch]

From a legal perspective, its still up in the air who is liable when your auto drive type system errors and someone "pays the price". I doubt Slate wants to spearhead those lawsuits.

 

[Daemoch]

Does slate use cameras to ID objects, or just 'sensors'? Cause Tesla is all about cameras and no sensors, which is just dumb, costs way more, and requires more power to process. If the slate is just doing the legal DOT basics, it could be just IR or sonic and meet the minimums which would cost less and require far less to implement. Like those round things in your bumper. And those WONT do any real drive assist realistically.

 

[Daemoch]

If a technology cant contend with literally ANY scenario it can ever come across, its not mature enough for the real world. full stop. thats just a real cold fact. Its why aerospace still uses human pilots after 50+ years of autopilots; its not ready to deal with "that one scenario" still and the lawsuits that would (rightfully) follow.

If someones Tesla 'auto' ran over your wife and you pulled them out of their car and beat them until they were unrecognizable by their own dog, then sued them and Tesla, I'd acquit you and then fine Tesla into oblivion to send a message to every other AI dev company out there. If a life is precious and beyond value, then why do we keep putting price tags on them? Otherwise...how much you want for your daughter? Asking for a friend.

If you want to use it, have fun. But if it literally steers you wrong and you spend the rest of your life eating through a tube, or in jail because you caused someone else to, I dont want to hear your cries of "I didn't know" "I'm so sorry" or "It always worked before". You rolled dice you didn't have to after all the warnings and lost. If driving is SOOOO tedious, dont buy a truck, take a taxi.

 

[Bayfire2441]

If a technology cant contend with literally ANY scenario it can ever come across, its not mature enough for the real world. full stop. thats just a real cold fact. Its why aerospace still uses human pilots after 50+ years of autopilots; its not ready to deal with "that one scenario" still and the lawsuits that would (rightfully) follow.

If someones Tesla 'auto' ran over your wife and you pulled them out of their car and beat them until they were unrecognizable by their own dog, then sued them and Tesla, I'd acquit you and then fine Tesla into oblivion to send a message to every other AI dev company out there. If a life is precious and beyond value, then why do we keep putting price tags on them? Otherwise...how much you want for your daughter? Asking for a friend.

If you want to use it, have fun. But if it literally steers you wrong and you spend the rest of your life eating through a tube, or in jail because you caused someone else to, I dont want to hear your cries of "I didn't know" "I'm so sorry" or "It always worked before". You rolled dice you didn't have to after all the warnings and lost. If driving is SOOOO tedious, dont buy a truck, take a taxi.

I don't think this is the way to think about driver assist at all. If they can do better in most scenarios than the average driver, I think thats good at the end of the day. No technology can ever be full proof and no person can ever be perfect. That's just life. Doesn't mean you can't take the better of the two options.

 

[E90400K]

I don't think this is the way to think about driver assist at all. If they can do better in most scenarios than the average driver, I think thats good at the end of the day. No technology can ever be full proof and no person can ever be perfect. That's just life. Doesn't mean you can't take the better of the two options.

But it should fail safe, not fail at fault. Fool proof is not the standard.

 

[Bayfire2441]

But it should fail safe, not fail at fault. Fool proof is not the standard.

It seems like Daemoch believes they should be fool proof. But, also, how do you make these systems fail safe? I see two routes, make it so the driver can take over without any issues. Which I believe every car maker does (with currently made models atleast). The other route I see is giving them more autonomy. More context hopefully means less failure. But I think by the nature of the world they operate in, you can't make a computer understand every scenario and I don't see how you make them fail safe without making them loose functionality. Like I said, if these assist and eventually autonomous driving can lower accident rates by a significant amount, I think its worth it to atleast try.

 

[E90400K]

It seems like Daemoch believes they should be fool proof. But, also, how do you make these systems fail safe? I see two routes, make it so the driver can take over without any issues. Which I believe every car maker does (with currently made models at least). The other route I see is giving them more autonomy. More context hopefully means less failure. But I think by the nature of the world they operate in, you can't make a computer understand every scenario and I don't see how you make them fail safe without making them loose functionality. Like I said, if these assist and eventually autonomous driving can lower accident rates by a significant amount, I think its worth it to at least try.

As I pointed out in a prior response of the group of Advanced Driver Assistance Systems (ADAS) are different than the Legacy Driver Assistance Systems (LDAS): Antilock Braking System, Traction Control, and Stability Control. The legacy systems provide passive control of the chassis because through computer assistance and specific mechanical hardware the brakes and engine torque in more precise manner than the driver can provide simply because the controls don't exist and really can't exist because the human only has two feet and two hands. All have limitations based the laws of physics. All fail safe. All can't overcome driver substantial driver error.

ADAS on the other hand take over safety responsibility from the driver, which is a vastly more in-depth mission statement. These systems have been developed in reaction to automakers building in distracted driver systems (i.e. the infotainment screen), which itself was added technology in attempt to control distracted driving from idiotic use of smart phones when driving. The concern here is as you stated, "you can't make a computer understand every scenario", which will eventually lead to drivers who do not have the capability to take over control of the vehicle when the ADAS reaches its limits because they never gained the skills through experience and practice.

@Daemoch mentioned air flight, which is quite insightful (and my area of experience with automation). While both commercial and general aviation aircraft have high levels of automation (along with redundant systems) the ultimate responsibility for control of the aircraft falls on the pilot(s). "Pilot(s)" because there is redundancy needed in the human system too. The difference between aviation and ground transportation is threefold, (1) pilots are constantly trained and retrained and tested for suitability to fly aircraft, (2) aircraft fly in 3-dimensional airspace (vs. 2-dimensional "groundarea"), and (3) aircraft fly in highly controlled airspace - for the most part. But the real foundation of air flight safety is pilot training (i.e. "better pilots").

To answer your question though. My answer is... No, we should demand "better drivers". I think Slate has the right idea, through the machine, make the driver more responsible for safety rather than less. That is the biggest value of the brand, in my opinion.

 

[Lynx]

If a technology cant contend with literally ANY scenario it can ever come across, its not mature enough for the real world. full stop. thats just a real cold fact. Its why aerospace still uses human pilots after 50+ years of autopilots; its not ready to deal with "that one scenario" still and the lawsuits that would (rightfully) follow.

If someones Tesla 'auto' ran over your wife and you pulled them out of their car and beat them until they were unrecognizable by their own dog, then sued them and Tesla, I'd acquit you and then fine Tesla into oblivion to send a message to every other AI dev company out there. If a life is precious and beyond value, then why do we keep putting price tags on them? Otherwise...how much you want for your daughter? Asking for a friend.

If you want to use it, have fun. But if it literally steers you wrong and you spend the rest of your life eating through a tube, or in jail because you caused someone else to, I dont want to hear your cries of "I didn't know" "I'm so sorry" or "It always worked before". You rolled dice you didn't have to after all the warnings and lost. If driving is SOOOO tedious, dont buy a truck, take a taxi.

The numbers don't lie.

Frankly, the data speaks for itself. Take a look at the data for the ONLY fully autonomous vehicle on the road currently that is actively providing rides without a human driver or human safety monitor: Waymo.

Compared to an average human driver over the same distance in their operating cities, Waymo had:

91 % Fewer serious injury or worse crashes
79% Fewer airbag deployment crashes
80% Fewer injury-causing crashes
92% Fewer pedestrian crashes with injuries
78% Fewer cyclist crashes with injuries
89% Fewer motorcycle crashes with injuries

This data is collected in the scale of per million miles. Of which Waymo has been studied over the course of 96M rider-only miles without a human driver. If you want to take a more in-depth look at the data I will link it here.

If you are genuinely concerned about the risk of harm to your loved ones from motor vehicle incidents, then you should look at the data for the leading causes of fatal accidents from National Highway Traffic Safety Administration (NHTSA) and the National Safety Council (NSC). The top three are:
1) Drunk driving
2) Speeding
3) Distracted driving

Autonomous vehicles such as Waymo don't drink and drive, they don't speed, and they don't get distracted or tired when driving. If you were to choose between ordering an autonomous Waymo to take you to take you to the airport or a human taxi driver, the statistically proven safer choice would be the autonomous Waymo. These numbers only speak to Waymo, not Tesla or other companies, but I think Waymo is the best example of the technology as they are the only one that is currently, right now as I write this, actively operating FULLY autonomous vehicles without any safety drivers or monitors behind the wheel or ready to take over.

If you were to pull someone out of their car and "beat them until they were unrecognizable", it would be the drunk, distracted, human that was speeding and killed your loved one. Human drivers are putting more of a "price tag" on human life then autonomous Waymo Drivers. If you genuinely seek a driving experience that is 100% full proof and makes driving completely 100% safe, then autonomous Waymo technology is far closer to your goal then human drivers. Regardless of your personal feelings toward the technology, the data speaks for itself.

 

[Dorbiman]

3) Distracted driving

This one is so big. The number of drivers I see on their phones while driving on my commute home is infuriating.