Transcript: Elon Musk’s press conference about Tesla Autopilot under v8.0 update [Part 5]

This is part 5 of our transcript of Elon Musk’s press conference about Tesla Autopilot under v8.0 update. You can read the part 1 and part 2 for Musk’s opening statement and part 3 is the first part of the Q&A followed by part 4.

Now we are at part 5:

Will Oremus – Slate

Thanks very much. this is a question I wanted to ask you for a while. You mentioned in some of the blog posts after Josh Brown’s death that the data showed you are safer driving a Tesla with Autopilot than manually driving a crash and the data that you cited was the number of fatalities per 100 million miles traveled. the number of fatalities is something that could change quite  substantially proportionally with one more death. So I wasn’t fully persuaded by that data point.

Do you have other data that you are relying on when you say that you are safer with Autopilot than without and are you looking maybe at like data about the number of injuries when people are using Autopilot and can you share any of that with us?

Elon Musk – Tesla CEO

Yeah. The full dataset is much more nuanced than fatalities. It should be said that we are now roughly at 200 million miles on Autopilot with only one confirmed fatality on Autopilot. that compares with a mortality every 89 million miles in the US and roughly one every fatality every 60 to 70 million miles in the rest of the world. and the numbers are counting the whole world. It should be compared to the whole world. Relative to whole world fatalities per mile it would appear that it’s roughly 3 times the number of miles.

But with each passing day, there are more miles accumulated. We are now adding roughly I think 1.5 million miles per day on Autopilot and that number is increasing as the Tesla fleet increases and as more people learn to use the Autopilot it has a greater percentage of miles driven. There’s more and more data coming all the time.

The thing that we think is really quite powerful is the fact that it tends to reduce the impact velocity. There might still be an impact, but if it decreases the impact velocity from something that would have been fatal or caused critical injuries to something where you walk away from the car – that’s enormously helpful and that’s a really significant difference.

But just looking at fatalities, it’s at least 2.5 to 3 times more miles per fatality and that number is growing every day. And the system is getting better. I think it’s really quite unequivocal that Autopilot improves safety and with this update it improves it even more.

I would estimate that with the improved fleet learning and software I think we will end up probably 3 times safer than a car that isn’t on Autopilot.That’s my guess. it’s not minor.

Jonathan Gitlin – Ars Technica

Hi Elon, before I wanted to try to ask you a question about Autopilot and sensors in 2015 [inaudible] The question I was actually going to make with radar then: Was to false positive right now [inaudible] Now whether or not you want to comment about if that was possible to achieve via Mobileye, does it mean that we might see lidar added to the system in the future?

Elon Musk – Tesla CEO

Sorry you blacked out for a second during your question, but we do not anticipate using lidar. Just to make it clear, lidar essentially is active photon generator in the visible spectrum – radar is active photon generation in essentially the radio spectrum. But lidar doesn’t penetrate intrusions so it does not penetrate rain, fog, dust and snow, whereas a radar does. Radar also bounces and lidar doesn’t bounce very well.

You can’t do the “look in front of the car in front of you” thing. So I think the obvious thing is to use radar and not use lidar.

What was the first part of your question? That blacked out unfortunately.

Jonathan Gitlin – Ars Technica

I was curious what was the false positive rate was in 2015 when radar was sufficient and what the false positive rate is down to  now?

Elon Musk – Tesla CEO

Which false positive rate are you talking about? You mean if we were to just use radar?

Jonathan Gitlin – Ars Technica

Yeah well evidently it wasn’t good enough last year when optical sensors alone were sufficient , but now you say that you can prove that the false positive rate to the point where you can rely on radar. So I was just curious what was that false positive rate before and low you managed to get it?

Elon Musk – Tesla CEO

I don’t actually have the numbers off-hand, but there’s really no way we could have used radar-only braking without improving the point cloud, the data associated with each point in that point cloud and very importantly the fleet learning to eliminate  aggressive braking for overhead signs and bridges  – any sort of metal thing or large dense things that is in front of the car but that is not actually dangerous because the road pitch will change.

It would have been probably false braking several times a year, which would have incredibly annoying, uncomfortable and dangerous.

Now we think that we can probably get that down to – I don’t know – most people will never experience one entire event with their car in more than 5 years or something like that.

Here you can read Part 6: Transcript: Elon Musk’s press conference about Tesla Autopilot under v8.0 update [Part 6]