My personal hunch is that fully autonomous self-driving tech is not theoretically possible under currently known computational models, because it implies many well-known NP-Hard problems. Self-driving companies are betting on the ability to find a heuristic/approximation that works "sufficiently well". But I strongly feel that the chasm that needs to be crossed to be "sufficiently good" is not one of magnitude (i.e.…
Do you believe human-driven tech is theoretically possible?
Current experiments and production data from human controlled vehicles have not been encouraging.
My personal hunch is that fully autonomous self-driving tech is not theoretically possible under currently known computational models, because it implies many well-known NP-Hard problems. Self-driving companies are betting on the ability to find a heuristic/approximation that works "sufficiently well". But I strongly feel that the chasm that needs to be crossed to be "sufficiently good" is not one of magnitude (i.e.…
That's basically what human driving is though, no? We consciously and unconsciously take our attention off the task all the time. It's not possible to drive fully alert of surroundings all the time and at least part of the time we are simply dead reckoning in fairly safe lanes of travel at constant speed and direction with minimal pedestrian and cross traffic. I could call it "controlled chaos", "luck", "planning" but there is some amount of unknown when moving a multi-thousand pound object around, and as speed increases so do the chances, ability to optimally correct, and severity of mistakes. It is very interesting and challenging to map the morality of the decisions of moving these machines onto automation.
Statistically speaking, I am curious to know the data on these areas: - number of miles driven per fatal accident on self-driving - number of miles driven per fatal accident on a human driver. It is important to know whether our machine is now doing worse/better than our counter-part and this means a lot for our self-driving initiatives.
1.15 per hundred million vehicle miles traveled by humans [1]. Uber only reached a million miles in September, with Waymo a little ahead at four million [2].
Note that this is far too little data from which to draw any conclusions, since the sample size for fatalities involving self driving cars is one and the number of miles driven is many orders of magnitude less.
Although it's comforting that this exact situation shouldn't happen again in an Uber autonomous car... there is no mechanism to share that learning with the other car companies. There seriously fucking needs to be a consortium for exactly this purpose: sharing system failures. Also my problem with this is that a human death is functionally treated as finding edge cases that are missing a unit test, and progressing th…
>There seriously fucking needs to be a consortium for exactly this purpose: sharing system failures. This implies they're using the same systems and underlying models. If one model hit a pedestrian because of a weakness in training data plus a sub-optimal model hyperparameter, and therefore was classified a pedestrian in that specific pose as some trash on the street, how do you share that conclusion with other compa…
I think there is likely a common need for a set of standard training models. This may hinder innovation for some time, but it's a cost we should accept when releasing a potentially dangerous technology to the public. It would have the added benefit of multiple companies contributing to a common self-driving standard which could accelerate its development.
That being said, when the first real cars were introduced to the world and later improved upon there were many more fatalities than we're likely to experience with self-driving technology.
My personal hunch is that fully autonomous self-driving tech is not theoretically possible under currently known computational models, because it implies many well-known NP-Hard problems. Self-driving companies are betting on the ability to find a heuristic/approximation that works "sufficiently well". But I strongly feel that the chasm that needs to be crossed to be "sufficiently good" is not one of magnitude (i.e.…
Do you believe human-driven tech is theoretically possible? Current experiments and production data from human controlled vehicles have not been encouraging.
This is what's going to happen. If you've ever seen a machine learning algorithm in action, this isn't surprising at all. Basically, they'll behave as expected some well known percentage of the time. But when they don't, the result will not be just a slight deviation from the normal algorithm, but a very unexpected one. So we will have overall a much smaller number of deaths caused by self driving cars, but ones that…
Very little of what goes into a current-generation self-driving car is based on machine learning [1]. The reason is exactly your point -- algorithmic approaches to self-driving are much safer and more predictable than a machine learning algorithms. Instead, LIDAR should exactly identify potential obstacles to the self-driving car on the road. The extent to which machine learning is used is to classify whether each ob…
Exactly, with LIDAR the logic isn't very tricky. if something in front, stop.
Statistically speaking, I am curious to know the data on these areas: - number of miles driven per fatal accident on self-driving - number of miles driven per fatal accident on a human driver. It is important to know whether our machine is now doing worse/better than our counter-part and this means a lot for our self-driving initiatives.
The numbers that I have seen indicate that humans have a fatal accident roughly once per hundred million miles, while we have one fatal accident for a self-driving car, with somewhere around ten million miles driven across all self-driving cars. I've heard Uber is around 3 million self-driving miles. So Uber would be 30x worse than humans.
Good to know! But I am more curious for the data cover every self-driving vehicles, not just about Uber. It could be Uber is doing worse, but my ask is whether we (as an industry) is still on a right direction for this "bold bet".
Very little of what goes into a current-generation self-driving car is based on machine learning [1]. The reason is exactly your point -- algorithmic approaches to self-driving are much safer and more predictable than a machine learning algorithms. Instead, LIDAR should exactly identify potential obstacles to the self-driving car on the road. The extent to which machine learning is used is to classify whether each ob…
Exactly, with LIDAR the logic isn't very tricky. if something in front, stop.
More than that - if something approaching from the side at intercept velocity, slow to avoid collision.
My personal hunch is that fully autonomous self-driving tech is not theoretically possible under currently known computational models, because it implies many well-known NP-Hard problems. Self-driving companies are betting on the ability to find a heuristic/approximation that works "sufficiently well". But I strongly feel that the chasm that needs to be crossed to be "sufficiently good" is not one of magnitude (i.e.…
Do you believe human-driven tech is theoretically possible? Current experiments and production data from human controlled vehicles have not been encouraging.
I added an edit to my comment: We don't have a computational/theoretical model for human consciousness. This is why it's called The Hard Problem of Consciousness.
Statistically speaking, I am curious to know the data on these areas: - number of miles driven per fatal accident on self-driving - number of miles driven per fatal accident on a human driver. It is important to know whether our machine is now doing worse/better than our counter-part and this means a lot for our self-driving initiatives.
The numbers that I have seen indicate that humans have a fatal accident roughly once per hundred million miles, while we have one fatal accident for a self-driving car, with somewhere around ten million miles driven across all self-driving cars. I've heard Uber is around 3 million self-driving miles. So Uber would be 30x worse than humans.
We also don't know how many fatal crashes were avoided due to human intervention (Uber has people inside the cars). It is almost certainly a whole lot worse than 30x at this stage.