Why self-driving does not work for me

As you know, Tesla and many other companies are pushing the idea of self-driving. “The car will drive itself; you only need to supervise it” – this is the message. “And later, your car will be smarter and can drive without any supervision”. But they all want you to buy now, and smartness will come later. Maybe.

But is it safe? Let me start with something simple. Even simple tools like adaptive cruise control and lane keep assist can fail, and you need to be ready for this case. Ideally, you need to know when it is safe to use them.

For example, imagine you are driving 75 miles per hour in the leftmost lane. To the left of you, there is a concrete barrier, and the road curves to the right. Imagine that lane-keep assist is disengaged. You need to analyze the situation, make a decision, and start executing it.

There is a very high chance that you will simply not have enough time to avoid a crash because you will have around 0.4-0.7 seconds before your car touches a concrete barrier.  There is also a chance that, due to time restrictions, you will turn right much more aggressively than needed and hit a car to your right or lose control completely.

As a result, I mostly use adaptive cruise control and lane keep assist at very low speeds or when the situation allows a lot of time to react. For example, on the highway when there are not many cars and the road is more or less straight. Of course, nothing stops you from using them at higher speeds, but I think it is not safe.

But let’s go to self-driving. Imagine you hired a driver who drives you in your car. Normally, this will allow you not to drive and allow you to do something else. For example, you can use a notebook or a phone. Or sleep, or just look around.

I think everybody can understand the trade-off here. You are spending some money, but you can do something productive in the time you usually spend driving. And I think everybody understands that if that driver damages another car or injures someone, that driver will be responsible, not you.

But imagine a situation where you are responsible for everything that driver does. If that driver damages a car, then you will pay for all damages. If that driver injures a person, you will go to jail.

Imagined? I think anybody who teaches another person how to drive immediately recognizes this feeling. You cannot relax; you have to constantly pay attention even when a car drives a very familiar road. It is also extremely exhausting. It is much, much harder than driving by yourself.

So if you are a responsible person, then you must constantly monitor your self-driving car and be ready to intervene at any second. But as I wrote above, it is extremely exhausting. It is way easier to drive yourself. You are driving a straight road. You know your car will continue to drive straight unless you touch the steering wheel.

But self-driving software can decide to turn left or right; it can brake or do something else. You must be prepared for everything. So in practice, many drivers just allow the self-driving car to drive, and they are not paying much attention. And sometimes they are not paying attention at all.

They lie to themselves that they are paying attention, but they are not. Then they get bored, and they have a temptation to check messages or email. As I wrote here, the technology almost works, and people start to trust it after some time.

They simply forget that they will be responsible for everything. The self-driving car will disengage 0.1 seconds before a crash. Obviously, the company will later state that, according to their telemetry, self-driving was not engaged at the moment of the crash. They will reiterate that the driver must pay attention at all times, and it looks like in this particular case the driver didn’t. We are so sorry.

What is most annoying is that they would not lie to you. It will be 100% true. And you cannot do anything in that 0.1 seconds because most of the time you are simply not ready. You must assess the situation, make a decision, and execute it. You need around 0.1 seconds just to move your leg to the pedal.

Of course, if you are driving, you can also make a mistake. You may also decide to check messages or email while driving. But you know that nobody drives a car while checking messages. A lot of people are still doing it, but at least they try to do it quickly.

With self-driving, you will spend way more time on the distraction because it seems that the car is doing fine. You feel less pressure, but when a car suddenly disengages self-driving, you will need at least a few seconds to do something about it. But most of the time you don’t have these seconds.

So, practically, in the current state, self-driving does not provide much value over adaptive cruise control and lane keep assist. It can help for long drives, but any sane person will not use it on city streets or in a busy traffic situation.

But if the company were responsible for anything self-driving does, it would be a completely different story. In this case, you hired a driver who drives you. You free yourself for other activities. It doesn’t mean that the driver will not make any mistakes, but at least you will not be responsible for them.

I personally still would not use it. At least in the first few years. Maybe even 10 years. But at least it sounds like a valid proposition. You pay money and get a service that can do something helpful to you.

Moreover, companies regularly overpromise what technology can do. For example, Tesla stated a few years ago that Hardware 3 would be enough for full self-driving. But now they are saying that it is not enough and its need to be upgraded to Hardware 4. As far as I know, the customer needs to pay for an upgrade unless they bought FSD.

Will Hardware 4 be enough? I don’t know, but I will not be surprised if we will need Hardware 5 a few years later to do Absolutely True And-Not-A-Lie-This-Time Full Self Driving.

Also, imagine the feeling of the people who bought Tesla with Hardware 2 and were also promised full self-driving. They paid $1000 to upgrade to Hardware 3, and now they need to upgrade and pay again. And they are not even sure that this is the last time. More importantly, they still didn’t have a product that they were promised.

Musk regularly promised that “soon” cars will do amazing things. Then time passed, and most of them are still not here. A lot of them were implemented many, many years later. And some of them still are not implemented. And those that are implemented are often very limited versions of what was promised.

In conclusion, I would like to state that, in my opinion, many companies make false promises about self-driving. I think that most companies right now do not provide a useful service to their customers and simply use them as free beta testers while these testers are paying for this privilege.

Moreover, these testers can go to jail for mistakes that self-driving software could make. And companies profit twice. The first time is when they get payment for self-driving technology, and the second time is when they use data collected by “testers” to improve the technology. It is a win-win for them, but very bad value for their customers.

I hope it helped someone.