|Self-driving cars are plagued with problems / Photo Credit: metamorworks (via Shutterstock)|
It’s 2020 and yet, those much-touted self-driving cars are still not on the road. A decade of waiting for these cars has come to a quiet end, according to Matt McFarland of business and financial news site CNN Business. Automakers and tech companies gave us a vision of a transportation utopia, investing billions in trying to make it a reality. In 2017, Tesla’s Elon Musk talked about autonomous cross-country trips and GM promised to make self-driving rides available in 2019. But these deadlines have come to an end with nothing but unfulfilled promises and humans still behind the wheel.
Matthew Johnson-Roberson, the co-founder of Refraction AI, explained, “Robotaxis have been three years away for probably the last five years. We're seeing a reckoning between what technologists understood to be the hard problems that were going to take us 10, 15, 20 years to solve and what was a hype cycle.” Various challenges facing the AV industry have slowed the development of self-driving cars and trucks. For one, it’s hard to prove that these vehicles are safe considering that sensors struggle in bad weather or they lack the human touch to safely navigate four-way stops and merging traffic.
A decade ago, self-driving cars were just mere science projects. However, the results from government-sponsored races impressed Google co-founders Sergey Brin and Larry Page, launching a self-driving project consisting of 20 people. Their self-driving cars were primitive but they made major strides this decade owing to breakthroughs in AI, particularly deep learning.
Ford’s self-driving arm, Argo AI, stated that advances are required for self-driving cars to navigate complex conditions with minimal computing power. “Until we're able to do so at scale, the visionary benefits that have been spelled out for society won't be achieved,” principal scientist Deva Ramanan argued. There are also other challenges, especially about AI. One issue with AI is blackboxes, making it difficult for developers to determine why the car drives as it does. When self-driving cars make a bad decision, it can be difficult to debug them. “We're between a rock and a hard place. Machine learning is right now not explainable or certifiable," said Raj Rajkumar, a Carnegie Mellon University professor.