The Elusive Quest for Fully Self-Driving Cars: A Deep Dive into Reality
As the race intensifies among tech giants like Waymo and Tesla to capture the potentially multi-trillion-dollar robotaxi market, the dream of fully autonomous vehicles remains just out of reach. In the latest episode of The Limit, viewers embark on a compelling exploration revealing that, despite significant advancements, true autonomy on public roads remains unattainable.
At the outset, the video emphasizes a fundamental truth: nothing on today’s roads is genuinely autonomous. While companies have made strides in developing self-driving technologies, the intricate challenges involved reveal that full autonomy is still a distant goal. This realization is particularly poignant upon examining the various technological approaches. Waymo employs a sophisticated blend of Lidar and camera systems, whereas others, like Tesla, primarily rely on cameras. This difference in methodology raises questions about the effectiveness and reliability of these systems under diverse conditions.
One of the most immediate challenges explored is the “school bus problem,” an issue that exposes the limitations of current AI systems. The unpredictable nature of human behavior, especially in complex urban environments—with children who may dart into traffic—poses risks that algorithms struggle to navigate. This leads to a broader discussion on the levels of automation, highlighting how the current capabilities barely scratch the surface of what is required for true self-driving.
Seasoned by weather’s wrath, the video delves into the challenges posed by snow and ice, which drastically complicate navigation. Whether it’s poor visibility or traction issues, harsh weather conditions highlight the nuanced scenarios that today’s autonomous vehicles are ill-equipped to handle. This sets the stage for the “long-tail problem,” where the rare but critical edge cases in driving scenarios can have dire consequences if not handled appropriately by the car’s AI.
Moreover, the video aptly discusses the human-to-car ratio, pointing out that for every fully autonomous vehicle, there are still numerous human-driven vehicles and unpredictable pedestrians. As long as human-driven cars fill the roads, achieving a seamless coexistence with autonomous vehicles is a significant hurdle. Further emphasizing this struggle is the segment that chronicles a ride stuck in a Waymo, where limitations of the system become glaringly evident when faced with unpredictable real-world scenarios.
The behind-the-scenes look at Waymo’s depot unveils how meticulously these vehicles are designed and maintained, yet it also showcases the level of human oversight still needed to ensure safety. Scaling up these technologies presents another formidable challenge, compounded by public resistance to fully embracing autonomous systems without proven safety records.
As the episode nears its conclusion, it contemplates the environments where full autonomy might first emerge. Likely candidates include controlled settings such as campuses or pre-defined routes, where the complexities of city driving are minimized.
In a thoughtful wrap-up, the video leaves viewers pondering the multifaceted barriers between the dream of fully self-driving cars and their realization, urging a candid reflection on the complexity of the technology, the unpredictability of human behavior, and the challenges that await on this exciting but fraught journey toward truly autonomous vehicles.
Watch the video by Business Insider
Video “Why Fully Self-Driving Cars Are Almost Impossible | The Limit” was uploaded on 03/21/2026 to Youtube Channel Business Insider















![Picking Up Our Malaysian-Made Truck Camper, But We Encounter Some Issues [S11-E3] Picking Up Our Malaysian-Made Truck Camper, But We Encounter Some Issues [S11-E3]](https://i.ytimg.com/vi/xxtQNrpDL4g/maxresdefault.jpg)





















Camera only driving can be accomplished at an acceptable margin of safety, that is an undeniable fact based on the number of cars on the road. The question Tesla is coming up against real hard is: By the time you are able to teach a computer to drive a car using a camera only system, will the cost of the additional sensors have fallen enough to make this a meaningful distinction? And on that question, I believe they will lose. By the time Tesla is able to perfect camera only self driving, the additional sensor suite will be such an insignificant added cost to the vehicle, and will make the job of self driving so much easier and safer, that it will seem like a stupid decision to run camera only.
Driving is social. Drivers wave each other through and signal to each other. Robots can’t do that.
Humans have been driving just 'fine' without Lidars and Radars for over 100 years. Sensors isn't the the limiting factor.
Activists are never right, they always take concerns with a grain of truth and blow it up to ridiculous proportions so that no can or should take them seriously.
How about this every time someone dies from these self driving cars, the governor of that state is executed, install a new governor and see what the new one does about these things.
Musk is a compulsively lying, sociopathic conman who has convinced tens of millions of people he is a genius, when he is clearly not. Anyone who thinks he is, after his ridiculous nonsense claiming a vacuum tunnelled roadway, hundreds of miles long, is easy, just like an air hockey table, is in a cult. He knows full well that Tesla cars are not fully able to self-drive safely and couldn't care less, as long as the money keeps coming in to keep the share price high. This is blatantly obvious when Tesla does all it can to fake demos, just like with their other products, and hide crash data, even going as far as to lie to authorities about data not existing when it does. Did I mention he is a sociopath?
I'd be interested to know how many accidents/fatalities the average human driver causes per million miles compared to Waymo.
Tesla's "camera only" is the reason why Tesla cars, in "self-driving" mode, crush motorcycles stopped behind other vehicles.
Stopped watching when you used bad AI slop to show an example where you could have hired an illustrator to create something presentable. DO BETTER.
Beta testing coming to a city near you!
Imagine the billions being put into this instead going to fix public transit.
2:15 wearing a kippa, loosing all credibility
I'm glad, as a human that drives, that I don't have a LIDAR… 2 eyes on a turret do the job.
Human drivers are not yet level 5 either…
That's what elon needs to do. Spend his own money. But first he needs to pay back our country for all the damage he has done during his doge age. In fact, just take all of his money. We don't need tessla. We can doge it!
I drive with FSD 95% of the time. My car never passes a school bus. My FSD is supervised. It is a much better driver than I am.
no one takes responsibility. until CEOs go to jail …… I'll walk.
Musk himself is expensive and unnecessary.
The thing with financial analysts did they predict. the 2008 crash (excuse the pun)
if lidars are so good why has luminar gone out of business and Waymo still driving pass school buses today and still have collisions. even with themselves!
Waymo is not level 4! don't kid yourselves by what they say. Even Tesal still say they are level 2 supervised.
even people don't drive in bad weather (snow, heavy rain, fog, etc) why do you expect a self driving car able to do that?
simulation does not work unless there is a brain to work out what to do in the scenario. Waymo cannot do this as it has no brain ( or a very small one!), Tesla has a much better and bigger brain and vision and best of all is a human with 2 cameras (eyes) and a massive brain to work out the edge cases. Even then humans still crashes due to lack of experience!
Yes, Waymo has kept remote drivers very quiet! wonder why?
lol. How is Tesla’s cost cutting method described as an “advantage”? What’s an advantage for Tesla’s bottom line doesn’t mean it’s an advantage for public safety .
We don't need this nonsense. Once all cars have this tech to level 5, then someone will be complaining about the Matrix controlling traffic, with cars driving you to an abattoir.
Phffft.
I don’t think we will ever get to a solid level 5 or 6. We all need to get off our damn cell phones to save more lives.
😂😂 you clown it will utilise satellite internet with multiple Blockchain to work with better technology and Blockchains that track where the car is along with sensors it will happen eventually it will just take time and better technology
Zero human traffic deaths is NOT a desirable outcome!
Humans need more natural selection, not less. Seat belts are a great example for this. Yes, it should be mandatory for car manufacturers to put them in and guarantee quality, but it must not be mandatory to use them, because humanity would be better of if those humans who don't use seat belts win a Darwin Award.
The problem with policy is to maximize opportunities to win Darwin Awards, while at the same time preventing these people from harming other people. How can this be done? Some examples:
– Optional seat belt usage. Brake hard or crash and you go through the windshield into a coffin.
– No more guardrails for cliff side roads. Loose control and you tumble down the cliff.
– More tree lined streets. Lose control of your vehicle and you wrap around a tree.
– Traffic lights at critical crossings could have wedge barriers in the stop line, that raise immediately once the light turns red. If you run a red light, you will be stopped by the wedge barrier.
– Obstacles on the road that need to driven around, just behind a tight turn. If you are speeding you smash into it.
– Fake road paintings on walls at T-intersections….
Just what I want: to buy a$70,000 car and then send it off to drive drunk college kids around.
This video deceptively showcases old teslas with very old self driving software against modern waymos.
Using old versions of FSD and set-up to fail tests by Dan O'Dowd is not representative of Tesla FSD today. Pull a manikin in front of any driver a few feet before the car gets there and the human driver would hit it. Tesla now always stops for school buses.
I like how full self driving is so stupid yet we are dumping a lot of money into it.
From an engineer perspective the technology ain't there and won't be there in a while. Also, it is just easier and cheaper to just optimise current transport systems.
Also read "Parents are using Waymo robotaxis as personal chauffeurs for their kids": https://bit.ly/4lzlkL8