The Philosophy of Autonomous Driving

 

[Dewg]

The Purpose

Why? Why do we need autonomous vehicles? If my car can drive itself, then why do I need a car?

Because sometimes you want to drive yourself, and sometimes you want be driven. This gives you the ability to do either.

 

[Tam]

There are many articles on this topic, and adding another exhaustive essay would only further stir the pot. Yet, I couldn't prevent my philosophy degree from spewing out all my ideas, and adding another perspective became a necessity. So sit back in your Tesla while charging and get your popcorn because this is gonna be long.
View attachment 996062




The Term Autonomous

The term autonomous is vague because the term AI itself is vague. The term AI is vague because defining intelligence is difficult. In order for a vehicle to count as autonomous, first it has to interact with its surrounding on its own account intelligently. If intelligent defined as the ability to solve problems, then our vehicle here is some what intelligent. However, intelligent is beyond the capability to solve problems. For example, cube solvers usually don't use intelligent to solve a cube. They use a mix of intuition and patterns of their past practices. But intuition is predicting the normal behavior of the external world without calculation. So should we add intuition to autonomous vehicle? Can we even do such a thing? And yet better, is it necessary to do that? Which comes to the question whether our autonomous vehicle should mimic human intelligence or should it be its own category of intelligence? Obviously human intelligence seems to fail sometimes, and on the other hand, human intelligence seems to prevail with most efficacy.

Whatever intelligence is, more is needed to determine whether a vehicle can interact with the outside world or not. A world full of determined and undetermined factors with many different variables that constantly changing. One concept is certain. That any form of intelligence used for autonomy must be adaptive to be ready for any purpose. But wait. What is the purpose of autonomous vehicle?


The Purpose

Why? Why do we need autonomous vehicles? If my car can drive itself, then why do I need a car?

Tesla said I could earn money while sleeping at home, and my car would autonomously pick up rides.

I was not interested in earning money from the Tesla Robotaxi Program. However, as I got older, my reaction skills got worse, so I needed a car that could drive itself better than I could, one that can auto park in a tiny garage without scratching any paint off.

If my car can drive itself, then why do I need a car

When a car can drive itself, it becomes my very own chauffeur without paying a salary to a human chauffeur.

It's convenient.

 

[RTPEV]

So should we add intuition to autonomous vehicle? Can we even do such a thing? And yet better, is it necessary to do that? Which comes to the question whether our autonomous vehicle should mimic human intelligence or should it be its own category of intelligence? Obviously human intelligence seems to fail sometimes, and on the other hand, human intelligence seems to prevail with most efficacy.

Yes, it is possible to add intuition, at least in an "artificial" way. AI is theoretically able to predict or intuit the future behavior of surrounding objects (pedestrians, animals, other vehicles) based on past behavior that it has "seen" in its training set (which in theory can be arbitrarily large). That's why I say it's artificial. It's not an inherent intuition. Rather it's based on past events it has seen that is far larger than any human could ever possibly have experienced. One could make the argument that human intuition is not really anything more than a prediction of future behavior based on what they have seen in the past.

Let's take an example. If you are on a two lane road with a car approaching in the other direction, your intuition is probably going to dictate that the car will remain in its lane. Even though it's possible that the car could veer into your lane, you probably ignore that possibility. But if you see the car veering or weaving, you become more alert and then intuit that maybe the car will come into your lane, so you take defensive measures, even if that has never occurred to you in the real world.

With an autonomous vehicle, with a large enough training set, it will have "seen" this behavior and the resultant departure of the vehicle from its lane, and could therefore take similar defensive actions. I am certainly not saying we are at that point, but it's theoretically possible. And given the limitations (and lapses) of our human intelligence, it's not inconceivable that the artificial intelligent control would surpass even that of most humans except for truly bizarre events that probably even humans wouldn't be able to intuit (a meteorite crashing down right in front of the car!)

If my car can drive itself, then why do I need a car?

You possibly don't. This is the big promise of autonomous vehicles, that we would no longer need to own and have the expense of a personal vehicle. If you own such a vehicle, it would likely be because you intend to use it to generate revenue, and thus the extreme valuations that Elon Musk places on FSD. I'm not sure how well thought out this is, however, because like anything else, there will eventually be an equilibrium between the number of autonomous vehicles on the streets and the potential revenue derived from them. At the two ends of the spectrum are the case where autonomous vehicles are numerous and thus very convenient (no more than a 2 minute wait for one to arrive when you decide you need to run to the store), but in this case the revenue earned by them to stay competitive with all the competing vehicles is low; and at the other end you have somewhat of a scarcity of vehicles, meaning revenue is high, but now you perhaps have to wait 15-30 minutes for a vehicle to show up.

And then what about a lot of the lifestyle activities that we have grown accustomed to? When I go to a concert or a sporting event, yes, it would be great to be able to just be dropped off at the gate, but what happens after the event when suddenly you have 15-80,000 people all exiting at the same exact time. First world problems for sure, but how quickly are people going to be willing to give up the convenience of being able to have their own vehicle in the adjacent parking lot where they can tailgate before and after and not have to wait in a huge line of people waiting for robotaxis?