Hardware in human eyes is actually pretty bad. Modern cameras can record and process more information in higher resolution and dynamic range.

The advantage is mainly in software. For example, I have a spot on my daily commute where I need to yield, but thick vegetation blocks the line of sight. I have to look for the signs of a moving car through the bush where there are some clearings, then wait to give the cars that I can't see the time to appear on the other side, if they exist, and proceed only then. I seriously doubt an image recognition system can do this.

A good set of stereoscopic cameras would tell the computer that those bicycles are not real, and even if they were, they are running at the speed limit so there's no need to break.

Radar would do an even better job. There's no reason why cars shouldn't have radars and drive better than humans, but yes, they should be able to solve that by vision alone.

> but yes, they should be able to solve that by vision alone.

Not really. A purely vision-based system probably needs to see this example in training labeled as "the back of a car" in order to differentiate it from reality.

A sensor-diverse system could detect the metal on the car and determine it is a car without ever having seen such an example during training.

And, this is just one edge case. The number of edge cases is infinite.

The point in the article is that a diverse sensor array is advantageous because it will cover more edge cases.

A purely vision based system will always have limitations compared to a system with more kinds of sensors.

> Not really. A purely vision-based system probably needs to see this example in training labeled as "the back of a car" in order to differentiate it from reality.

Have you take a look at depth based feature recognition? It will automatically recognize everything as a car, and will tell the error of the 2D feature extraction every time.

It's also extremely CPU intensive, so people don't apply it as often. It's a clear example of AI being hold down due to limitations in computer power.

Why would you want to guess at a 3d mapping from a 2d image when you can get 3d data directly from a sensor, like lidar?

Getting 3d from 2d images is certainly interesting and will have its applications. It's less accurate and more computationally demanding than using 3d input directly, however. Cars can and should have more than 2d input.

The point of the article is that using a diverse array of sensors in cars will help the algorithm make better predictions.

There are many edge cases out there that aren't covered,and we shouldn't expect engineers to plan for them all. What we want to do is throw a lot of raw input at an algorithm and run it through a ton of simulations. All steps of this are non-trivial, but with more data, it is easier.

Well, in this case we have a clean slate, though, so we can decide what is optimal. It may be that LIDAR is more effective than a purely binocular approach to detecting distance.

You could. But for that you need a good model of the world. Which I doubt any self-driving car AI has.

Most drivers also use binaural hearing.

Humans are kind of shitty at driving though.

You have to define "shitty" in this context. I agree that there are a lot of bad drivers out there. But even with those bad drivers, people driving while texting/intoxicated, and with older vehicles that don't have all the latest safety and assistive driving features, the fatality rate in the US is still something like 1 per 100 million miles. There are a lot of auto-related deaths in the US but there are a lot of miles driven.

Given modern safety standards in vehicles, I don't think fatalities are a good measure of how good humans are at driving.

It's news when a self-driving car runs a red light, but pretty normal for humans.

That may be so, but they're not often confused by bikes painted on the back of a van either.

I think this is only partially true. When I saw the image I was actually confused by why the outlines were wrong. In real life though I wouldn't be presented with a single image, but rather high res video.