Disruption Explained: Tony Seba - does this change your view of Elon?

[Brando]

rEVolution 2018 Amsterdam

 

[KJD]

Thanks for posting that Brando. That was an interesting presentation.

Elon has done some amazing things with Tesla as a car company and I think Tony Seba pointed that out pretty well. It is also apparent that he thinks that Waymo is way ahead with the self driving technology.

I don't know anything about Lidar or Waymo, but I will say that the FSD option on my model 3 has so far been a disappointment. Maybe Tony is correct on this one. What do you think ?

 

[Brando]

Thanks for posting that Brando. That was an interesting presentation.

Elon has done some amazing things with Tesla as a car company and I think Tony Seba pointed that out pretty well. It is also apparent that he thinks that Waymo is way ahead with the self driving technology.

I don't know anything about Lidar or Waymo, but I will say that the FSD option on my model 3 has so far been a disappointment. Maybe Tony is correct on this one. What do you think ?

Elon has it nailed.

Lidar uses photons (light even if "laser"). Not much different than regular light.
"image recognition and what to do" you solve for visual light, then you must solve for Lidar (twice the work).
IF the two different sensors don't agree - what do you do? So added complexity, seldom a good thing.

Extreme case you could use Radar, Lidar, camera and sonar when they don't all agree, what do you do?
Very complex to blend all these different views of the world and then decide what to do.
As Elon says, finding driveable space is the goal. GPS or Maps can't handle construction or broken down cars so not much help.
Watch this video - long & detailed - rewind often to understand. Totally worth the effort. look up words you don't understand.

Certainly not an easy problem. Big Data is key for Machine Learning - no one has anywhere near the data Tesla has gathered.
Elon always the optimist - and his time lines aggressive - when? your guess as good as mine. Elon's better. 2 years? seems certainly 5 years?
RoboTaxi? a new problem - how many "regulations" do you see a taxi break just picking up/dropping off passengers? Ask Lyft/Uber drivers.
illegal/double parking - stopping in the road ...

You are seeing better than most, how much work needs to be done. REMEMBER still a work in progress. Best to perhaps think of it as a co-pilot to "help" you drive - reminder texting drivers worse than drunk drivers and cause about 8-10,000 extra deaths each years since 2014 - far from RoboTaxi at this point, right?

Sorry, I don't remember Tony's timeline. He is awfully good, but it is hard to predict, especially the future of things never seen/done before.
Interesting times. Never thought I'd see coffee shops or brewers in SLC. How about a winery and grape growing is that started yet?
And is Provo still caffeine and alcohol free?? I miss Utah Coal and Lumber in Park City. If you ever get chance to go to film festival that was worth doing. I enjoyed it.

enjoy, make friends, help them, they may return the favor - can't have too many friends and remember we are all far from perfect so try to understand our short comings, empathy and forgiveness always helps. You can't control others but you can control HOW you feel/react to others. good luck

Trust your co-pilot but always verify.

 

[Buggle]

Elon has it nailed.

Lidar uses photons (light even if "laser"). Not much different than regular light.
"image recognition and what to do" you solve for visual light, then you must solve for Lidar (twice the work).
IF the two different sensors don't agree - what do you do? So added complexity, seldom a good thing.

Extreme case you could use Radar, Lidar, camera and sonar when they don't all agree, what do you do?
Very complex to blend all these different views of the world and then decide what to do.
As Elon says, finding driveable space is the goal. GPS or Maps can't handle construction or broken down cars so not much help.
Watch this video - long & detailed - rewind often to understand. Totally worth the effort. look up words you don't understand.

Certainly not an easy problem. Big Data is key for Machine Learning - no one has anywhere near the data Tesla has gathered.
Elon always the optimist - and his time lines aggressive - when? your guess as good as mine. Elon's better. 2 years? seems certainly 5 years?
RoboTaxi? a new problem - how many "regulations" do you see a taxi break just picking up/dropping off passengers? Ask Lyft/Uber drivers.
illegal/double parking - stopping in the road ...

You are seeing better than most, how much work needs to be done. REMEMBER still a work in progress. Best to perhaps think of it as a co-pilot to "help" you drive - reminder texting drivers worse than drunk drivers and cause about 8-10,000 extra deaths each years since 2014 - far from RoboTaxi at this point, right?

Sorry, I don't remember Tony's timeline. He is awfully good, but it is hard to predict, especially the future of things never seen/done before.
Interesting times. Never thought I'd see coffee shops or brewers in SLC. How about a winery and grape growing is that started yet?
And is Provo still caffeine and alcohol free?? I miss Utah Coal and Lumber in Park City. If you ever get chance to go to film festival that was worth doing. I enjoyed it.

enjoy, make friends, help them, they may return the favor - can't have too many friends and remember we are all far from perfect so try to understand our short comings, empathy and forgiveness always helps. You can't control others but you can control HOW you feel/react to others. good luck

Trust your co-pilot but always verify.

I'm skeptical on FSD, here's why -- Elon always goes on about how we drive with two cameras (our eyes). But this isn't really correct, we drive with two cameras connected to our brain. How many times have you been out driving, and you see something off in the distance, but are unable to determine what it is until you get closer, it comes better into view so your brain can make a better determination as to what it is, or you keep watching it, allowing your eyes and brain to use depth perception to help figure it out, then your brain makes a decision as to what to do only once you've determined what it is. I'm not in any way any kind of AI or computer expert, but I'll try to give an example.

Say you're driving along the freeway, there's a guardrail on the right. at the end of the guardrail, you see a rectangular black outline in the distance. You don't have enough information for your brain to know what it is, but you'll keep an eye on it, suspicious it could be a cop. Say you get closer to it, you're driving in the right lane, and it turns out it is a cop, and it turns on it's overheads as you approach. FSD won't react in any way to that. A human will look at it, become cautious that it's about to enter traffic in a hurry, and move over and give it some space.

This example I'm sure is by no means perfect, but is the best I can think of off the top of my head. I really do not think Elon is correct when he says we use only cameras to drive. We use cameras AND judgement. I understand you can train AI's for most situations through experience, but anyone who has ever been in an accident before can tell you.... there are many different edge cases, and most of them are different. I think we're a long, long way from AI being able to perform anywhere near as capably as a human brain. I understand that AI will be a better driver, and therefore will be able to avoid most fender benders caused cause someone who was looking at their phone, drinking coffee, following too close, etc, but I'm not sure that's going to make it good enough. This will avoid 90% of accidents, but it won't help with that 10%, which contains a good portion of the really serious ones, in which the brain plays a key role in avoidance.

 

[Brando]

I'm skeptical on FSD, here's why -- Elon always goes on about how we drive with two cameras (our eyes). But this isn't really correct, we drive with two cameras connected to our brain. How many times have you been out driving, and you see something off in the distance, but are unable to determine what it is until you get closer, it comes better into view so your brain can make a better determination as to what it is, or you keep watching it, allowing your eyes and brain to use depth perception to help figure it out, then your brain makes a decision as to what to do only once you've determined what it is. I'm not in any way any kind of AI or computer expert, but I'll try to give an example.

Say you're driving along the freeway, there's a guardrail on the right. at the end of the guardrail, you see a rectangular black outline in the distance. You don't have enough information for your brain to know what it is, but you'll keep an eye on it, suspicious it could be a cop. Say you get closer to it, you're driving in the right lane, and it turns out it is a cop, and it turns on it's overheads as you approach. FSD won't react in any way to that. A human will look at it, become cautious that it's about to enter traffic in a hurry, and move over and give it some space.

This example I'm sure is by no means perfect, but is the best I can think of off the top of my head. I really do not think Elon is correct when he says we use only cameras to drive. We use cameras AND judgement. I understand you can train AI's for most situations through experience, but anyone who has ever been in an accident before can tell you.... there are many different edge cases, and most of them are different. I think we're a long, long way from AI being able to perform anywhere near as capably as a human brain. I understand that AI will be a better driver, and therefore will be able to avoid most fender benders caused cause someone who was looking at their phone, drinking coffee, following too close, etc, but I'm not sure that's going to make it good enough. This will avoid 90% of accidents, but it won't help with that 10%, which contains a good portion of the really serious ones, in which the brain plays a key role in avoidance.

watch the video - of course NO ONE is doing it yet. Your points well taken, of course.