[intended]

Self-driving is never going to happen, and its also sitting in some sort of informational blind spot for the people working on it - I have no idea why.

There is no such thing as "driving" there is no physical force, or particle. There is no force preventing you from driving in the opposite direction of traffic, or through glass panes.

"Driving" is entirely a social phenomnon, the confluence of societal self impositions and engineering.

If you have a car on fire in front of you, you will need to reverse in the wrong direction of traffic.

In many countries - you have to regularly deal with drivers going full tilt, on the wrong side of the road.

Or You have to deal with theft, and people trying to rob you at every red light.

Im underscoring that this is a social issue. You would need to create models for each country and region, to truly improve self driving.

Self driving assumes a far narrower problem space than reality gives a fig for.

Self driving theory currently works in the same way any theory that assumes spherical cows works.

[esperent]

> In many countries - you have to regularly deal with drivers going full tilt, on the wrong side of the road.

I don't agree with most of your comment but this point is worth examining. I think it's not an argument against self driving for a few reasons:

1. There's lots of places in the world where people driving on the wrong side of the road is uncommon. We can start with self driving there. You can apply this to many other situations as well - AI can't handle ice yet? Well, let's start with non-icy roads. Even when you apply all the stipulations like these that you need, you'll still be left with a large enough percentage of the world to make self driving useful. Especially since a lot of the places that are suitable will be rich cities in developed countries.

2. Driving habits change. Thailand is a good example of this, it's a country in transition from the "developing country driving style" to the "developed country driving style", for want of better terms. Driving there 15 years ago was an extremely different, far noiser, far more dangerous experience than driving there now. It's still got a long way to go, of course.

3. If self driving becomes the norm, well then, problem solved. We don't even have to get to FSD. Partial but always on assisted driving that nags you whenever you drive on the wrong side of the road or go over the speed limit would probably be enough to cause a shift if most people have it.

[intended]

Therein lies the rub

Movement, is animal like - perceiving self, environment and understanding movement through it. - Movement is universal. Its the application of physics.

Thats not driving.

Driving is a social construct. It is the application of physics while navigating a social world.

Driving is observing the law, observing social constructs (that differ regionally), adapting to new constructs based on location and environment.

Thats the blind spot for self driving proponents. They conflate the two things, but talk primarily about the first.

As a result, you will never get self driving - the assumptions are wrong.

Let me put it this way - you get self driving when a car decides its best course of action is to reverse in traffic, because it perceives a tsunami coming from ahead of it.

[esperent]

> Driving is a social construct. It is the application of physics while navigating a social world.

Interesting way of looking at it. In the countries where people tend to drive at high speed on the wrong side of the road, I'm inclined to agree with you.

In countries where people follow the rules of the road, I think it's the opposite. There, it's primates who evolved to follow and react to a complex social and physical environment being forced to follow a simple and rather tedious set of rules, which require constant vigilant attention. This is something we humans are not good at - we're designed for short bursts of attention in situations where rules need to be interpreted intuitively, not driving for three hours while strictly following the rules of the road.

> observing social constructs (that differ regionally)

Well, the social constructs around driving also different temporally and are changing all the time. Self driving, even partially implemented, is sure to have profound effects on these cultures all over the world.

>perceives a tsunami coming from ahead of it.

In extreme cases like tsunami, floods, or fires, there are solutions. In the short (and probably medium) term the cars will still have human controls, so the human can instantly take over. In the long term, well maybe there can be a "panic button" that allows a human to shout voice controls. Either way, it's just another solvable technical issue, and probably much easier than many other issues that need to be solved.

BTW it's not like researchers haven't considered to these things

https://www.thecarconnection.com/news/1122534_self-driving-c...

[jstx1]

I don't understand the leap from "it's difficult and there are many special circumstances" to "it's never going to happen". I don't get why you think that any of these problems are unsolvable given enough time.

[abainbridge]

Here's one way to get to the "it's never going to happen" outcome:

The grand-parent comment argues that we need AI that works in a much broader set of circumstances to "solve" self driving. In particular it would need to understand how other humans would react to its actions in novel situations. That is approximately a description of having a Theory of Mind. Some argue that you can't have a Theory of Mind without being conscious. We might ban captive conciousnesses for ethical reasons.

I think you could invent lots of other scenarios that yield the "never going to happen" outcome. They probably all sound ludicrous, because having an AI that understands the workings of human minds sounds ludicrous (and frightening).

[Animats]

Actually, I'm not saying that. I'm saying that self-driving is nearing success, and that it took a while. LLMs are at the beginning of that process.

[neuronic]

To me this is a technocratic stance - believing that technology will always find a solution and absolve us of all problems.

Maybe. Maybe not. What I read from his post is much closer in spirit to "we cannot simulate real physics down to the atom" when modeling protein folding so what we do instead is building models (good enough estimations) and AI approaches to recognize complex (but still top level) patterns.

These reduced views on reality help our problem space to a large degree but they can never account for the full scope of the physical reality and they largely work on assumptions which might be proven wrong at any point in time.

[intended]

Thats not the point.

The point is people who are proponents of self driving have a blind spot. Its construed as a code problem, ignoring entirely the massive social aspect of human behavior.

Driving isn't locomotion. You can't assume that the people around you are going to be reasonable, sane or predictable. More data doesn't fix that, unless you start building a "civilzed behavior" model and then add that to your movement model.

I really want to see how honest people are when they build a fair representation of human civilization, warts and all.

Self driving as has been defined assumes some absurd things about the world in which humans live. No one is going to buy a car which wont know to run when they are about to get robbed.

To be blunt, the blind spot assumes you live in America (and probably california), not that you live in Brazil or India or Egypt.

[salawat]

Because no matter how much time you give it, there are social problems that need to be solved first; and said social problems are not the kind that are amenable to mere technical solutions.

Literally the case of: If the technically perfect implementation existed tomorrow, we still would not be ready to flip the switch because of how drastic the reorg of societal norms would be. It'd be a complete refactor.

[ryanwaggoner]

This comment is pretty funny in the context of a world where there are currently cars driving millions of miles without human intervention in a variety of conditions. I don't think anyone cares whether it meets your personal definition of "self driving".

Reminds of the Chinese proverb: “Man who say it cannot be done should not interrupt man doing it.”

[intended]

Cmon, - you are the one misrepresenting the actual definition of self driving - aka Level 5.

No vehicle is at level 5.

Calling it “my definition” and then switching the actual definition is unfair.

[ryanwaggoner]

Who decided that only level 5 qualifies as "self-driving"?

[intended]

This comment is pretty funny in the context of someone starting out by claiming no cares “whether it meets your personal definition of "self driving".”

[ryanwaggoner]

No, you said that I changed the “actual definition” of self-driving from Level 5 to something else. I’m asking what makes Level 5 the “actual definition” and not just an arbitrary point that you’ve personally decided means self-driving?

[tim333]

I guess like heavier than air powered flight will never happen. I mean there would be so many problems.

[pferde]

Indeed, it would probably require the invention of a true Artificial Intelligence, and of high quality as well. Not the glorified autocomplete the scammers are trying to pass as AI now, but the real thing that can sense and understand the world around it.

[gremlinsinc]

it'll happen but the car won't really be self driving, I think a literal ai robot cheaufer is more likely. AGI will be able to do everything a human can do, including driving. We're IMHO less than a decade, if not less than half a decade from that.

If you don't believe me name a scatter chart showing ai papers since 2017. Notice how this year dwarfs all previous years and last year blew the water out of the previous ones but the ones before generative art models were a bit slower and more regular.

If we can perfect ai agents and automated LLMs, we could create personified agents of varying backgrounds in a virtual computer lab tasked with doing ai research 24/7. We could put 60 ai scientists in the lab and just watch them work.

Maybe the lab is vr such that human scientists could enter and collaborate with ai counter parts. imagine applying this to cancer research, etc.

When an LLM can train itself, program children(next version), and submit orders or plans for new hardware fabs to a factory to breach current limitations, then we basically have von Neumann probes that multiply, grow, learn, and don't need human intervention.

[simiones]

LLMs are nowhere near being able to research other, better LLMs. It's not even known if existing LLMs are at 1% or at 99.99% of what an LLM can in principle achieve.