Do not forget the driver’s education. Soon, driving cars is more effective than teaching humans.
Artificial intelligence might be inaccessible to the untrained eye but it’s in action on the city roads throughout Shanghai and Suzhou and Suzhou, where automobiles are teaching each other to navigate themselves, thanks to an alliance with SAIC Motor, the largest automaker in China, and Suzhou-based artificial-intelligence company Momenta.
The companies collectively employed 60 robotaxis across the two megacities during the time between December 2021 and March 2022 in order to validate technology that could be utilized in cars within the next few years. Through an app for smartphones, the city’s residents in China performed their daily chores by calling robot taxis with human operators who are able to control the vehicle in the event of an emergency. Momenta states that 80 percent of users have utilized this app at least once following their initial trial and proving the “great potential” of the service.
Momenta is seeking to extend the technology it has developed to General Motors, Mercedes-Benz, and numerous other automakers with the goal of launching passenger vehicles in the next few years. In the beginning, Momenta CEO Cao Xudong states that he will need to collect an equivalent to 100 million miles of driving information.
Many firms across the US and across the globe like Cruise as well as Argo AI, are developing new technology for autonomous vehicles. Momenta’s program could come into effect sooner. Similar to other methods, it’s a machine-learning system that records and monitors millions of movements performed by humans, but instead of using these movements as the foundation for establishing driving rules, which are hard-coded Momenta allows data to send instructions to the other vehicles directly.
Furthermore, the company enjoys two major advantages over competitors having access to China’s high-definition road maps as well as strong alliances with automakers who already gather information from a myriad of vehicles that are on the road. General Motors, Mercedes-Benz, Toyota auto parts manufacturer Bosch and many others invest more than $1 billion into Momenta to help accelerate the development of autonomous vehicles.
“The challenge of solving scalable autonomy is huge,” Xudong says to Fortune. “When customers see that we are capable of developing algorithms for mass production programs, they understand our capabilities.”
The classical rules-based method of machine learning instructs vehicles to respond to stimuli by performing a specific action. “If X happens, you do Y,” says Ram Vasudevan, associate professor of mechanical engineering at the University of Michigan. “If the car’s camera reads a stop sign, you stop.” The software is programmed by developers using rules for various movements for example, such as how to change lanes or move around a vehicle that is not moving. If you turn left, for instance, the vehicle can be set to recognize the direction and slow down to match the path.
But anyone who’s been in a car knows that the traffic situation isn’t always as straightforward. Let’s look at an example of a kid running down the street to chase an object. It is up to the driver to decide whether to stay or change lanes. Numerous companies have programmed a lane-change move, However, Momenta’s data allows the car to consider other factors, such as the weather and timing of the day.
That’s why Momenta has abandoned rules-based algorithms in order to build cars completely from the ground up. “What Momenta is doing is truly artificial intelligence,” Vasudevan says.
The HTML0 “data-driven,” end-to-end learning approach is much more difficult and costly to develop. The millions of miles that partners track in the roadways aid cars in learning to function in different situations. The only way to collect such a large amount of data is through collaboration with automakers that are making billions in the manufacturing of their autonomous vehicles.
The benefit of Momenta’s method is that it’s able to deal with more complex and complex scenarios as well as nuances. “How do you know that you’ve preprogrammed all the rules out there?” Vasudevan declares. “The problem is that you’ll need lots of information labeled to create the right rules for each driving scenario. After all that there’s a chance that you’ll run into something that you’ve never included in your computerized stack. “
It’s true that the market demand for autonomous automobiles is growing quickly. In 2025 Momenta estimates that 60 million cars will be equipped with level 2 and 3 assistance driver technology. These will be the first step toward fully autonomous vehicles.
The U.S., federal regulators have declared that automakers are not obliged to outfit completely autonomous vehicles equipped with steering wheels, brake pedals, or other controls that are manual for driving. This opens the possibility for American automobile manufacturers to try their cars in public roadways.
the Cruise division of GM’s owned Cruise as well as Alphabet’s Waymo were granted permission from the California Department of Motor Vehicles to operate robotaxis on roads that are public in certain conditions. Tesla CEO Elon Musk disclosed during the quarter-end earnings meeting in April that Tesla is planning to develop robotaxis that do not have pedals or steering wheels by 2024. “I think that that really will be a massive driver of Tesla’s growth,” Musk declared.