( See ,” Artificial intelligence faces serious roadblocks,” Asia Times, July 18, 2017 ) The viability of autonomous vehicles for general use in public traffic has long been questioned. Years later, after spending billions of dollars, we are faced with the fact that the technology still falls short of standards for dependable public traffic use.
When the autonomous-driving division of General Motors, Cruise, was shut down by the California Department of Motor Vehicles next fortnight, this shortcoming came to light. The expulsion was brought on by an event on October 2 in which a human-operated aircraft struck and pushed the feminine pedestrian into the course of an uncontrollable Cruise car.
She was dragged on 6 meters after the driverless vehicle collided with her, came to a stop, and then tried to pull over to the side of the street. Undoubtedly, the demands of this circumstance were not met by the vehicle’s computer system.
It’s not because there has n’t been enough technological investment; through extensive road operation, billions of dollars have gone into developing autonomous vehicles with sophisticated sensors and sophisticated software.
There was a perception that these cars were getting ready to handle the complexity of regular highway customers. In addition to GM’s defeat, the self-driving vehicle manufacturer TuSimple  Holdings is liquidating, wiping out billions of dollars in market value.
Despite these ongoing setbacks, optimists continue to believe that artificial intelligence ( AI ) software will eventually enable vehicles with better driving performance, potentially saving countless lives by reducing human-caused accidents.
Prior to the disqualification, GM had estimated that its autonomous taxi business would generate US$ 50 billion in revenue by 2030. The subsidiary was valued at$ 30 billion in a 2021 funding round, reflecting this upbeat forecast.
Why is this software so difficult, given the wonders brought about by new AI software technology?
These self-driving cars stand out from other robots made for commercial or industrial settings in a significant way. Regular robots are made to perform predetermined programming tasks, like soldering or manufacturing painting.
To reduce harm to the creation line, the robot stops or engages in predetermined techniques if expected events interfere with the assigned process. These default-programmed techniques can be confidently carried out in standard production environments.
Then think about a self-driving car. Mechanical performance can become programmed because the planned activities are constrained if it is intended to operate in a controlled environment where things are repetitive, such as closed streets or restricted tracks. For example, driverless trains on inter-airport rail lines run properly because track access is restricted. The cameras may pick up on any obstructions and stop the train if they occur.
However, an autonomous car like the GM vehicle, which can move easily through available traffic, must function properly under virtually limitless circumstances. In essence, it is anticipated that the car will mimic human knowledge. Despite assertions that these cars will surpass people drivers, human knowledge continues to be important due to its special abilities.
Customers accidents are by nature arbitrary occurrences that differ from one another in some ways. Rapid considering and access to information that the vehicle robot cannot see, such as external data that automated sensors might not be able to pick up, are necessary to handle such situations. No amount of training can get the car ready to handle any kind of transportation situation.
To” coach” these vehicles, thousands of kilometers of traffic driving have been used, and billions of dollars have gone toward software upgrades. However, they occasionally run into situations where the” trained” conditions do n’t match reality and the vehicles perform worse than what would be expected of human drivers.
Will additional investment overcome these mechanical restrictions, and will autonomous vehicles outperform people drivers? are the remaining questions. They’ll probably find their specialty in handled settings, in my opinion. Taking the place of common people driving? Hmm.
Henry Kressel is a long-time private equity investment in tech companies as well as an innovator, technologist, and writer.