What are the challenges faced by self-driving vehicles?
Every year, Elon Musk seems to suggest that autonomous vehicles are at the door, and that they will be used to power vehicles traveling in small tunnels at high speed as well as rovers on the surface of Mars. But with all his positive intentions, the truth is that The concept of a self-driving car is still very far from reality.
It is true that there are test vehicles that are capable of navigating around the city center and can reliably detect and avoid obstacles, but in most cases, these environments are highly controlled by developers, eliminating many unknowns. A classic example of this can be seen in California with the self-driving taxi service operated by Waymo. Vehicles operated by Waymo are limited to specific well-documented and mapped areas, and a video presentation by YouTuber Veritasium showed that construction signs requiring vehicles to merge and move around confused the Waymo vehicle so much that it had to stop and ask for help.
But why is self-driving a difficult challenge to solve?
Simply, Driving is an incredibly complex task that requires a combination of visual processing, understanding and interaction. Furthermore, decisions must be made in rapidly changing situations that the driver has not previously taken into account, and many factors such as environmental conditions and road changes must be taken into account.
Self-driving cars cannot be powered by basic computational algorithms that run analytical models for every conceivable event. Instead, self-driving vehicles undoubtedly require the use of artificial intelligence that is trained to recognize the most important situations while also learning how to handle evolving ones.
For example, a self-driving AI system does not need to be told precisely what to do in each situation because it can understand that if something gets in its way, it must take safe measures to either turn the car around or move carefully. obstacle. Furthermore, the AI must also be able to read road signs and recognize that map data may be outdated (eg new intersections, roundabouts, and roads).
Creating such AI requires huge amounts of data; This is where we see one of the biggest obstacles; real world data. Creating a self-driving car that collides with other cars might be great for learning AI but is not entirely appropriate as a method for collecting data. Connecting human-driven cars to the internet and streaming live data is another option, but this also presents challenges such as connectivity and privacy.
In general, the lack of data and the current state of artificial intelligence technology is unable to produce a self-driving car that is safe to operate.
How does Tesla use customers to train their AI models?
Tesla is best known for its development of electric cars, but while most people see Tesla as a car manufacturer, what Tesla really means, arguably, is a data company. Every Tesla vehicle comes with an array of sensors, cameras and data loggers on board that streams this data to Tesla and serves the sole purpose of training self-driving AI. With every mile that Tesla cars drive, the AI models that Tesla develops are getting improved, This use of customer data creates a feedback loop that helps improve future versions of the Tesla Autopilot software.
At the same time, using firmware updates allows customers to upgrade their vehicle’s capabilities over time, and those who sign up for beta testing programs can access features still in development. This use of beta testing allows Tesla to get real-world data that would otherwise have been very difficult to obtain, giving Tesla an edge over its self-driving competitors.
Is Tesla Using Customers As “Guinea Pigs” Dangerous?
In most circumstances, using customers to test beta versions of software before it comes out can be an excellent way to test the waters. Possible bugs that may cause grief can be eliminated, and customer feedback can help improve existing services and add new features. However, in the case of autonomous driving, giving customers access to beta versions of software in order to collect data and improve performance is not only dangerous, but also unethical.
newly, Tesla gives early access to a full autonomous driving system for select beta customersVideos of drivers using the software to navigate around mountains and other dangerous environments are already being posted. While drivers have to remain alert and ready to take charge if something goes wrong, it’s very likely that those testing the new features don’t fully appreciate that so-called “full autonomous driving” is closer to adaptive cruise control with an automatic Maintain the lane of an actual self-driving vehicle.
Worse, those who test Full self-driving software may also not realize that the software is a trial version And as such, it can easily get into mistakes. This is especially dangerous for other drivers who may end up in a collision due to a mistake made by the beta program.
Simply put, the “full autonomous driving” beta will undoubtedly encourage illegal driving and put the lives of drivers, passengers and other vehicles at risk. If Tesla wants to test self-driving systems, it must do so in company-owned vehicles in environments that do not put others at risk.