Tesla’s layoffs hit Autopilot team as AI develops

Tesla (TSLA) is increasingly turning to machines rather than humans as it attempts to develop autonomous vehicles.

As part of Tesla’s plans to cut 10% of salaried staff, the company has laid off a significant number of its data annotation specialists. These specialists do grunt work that is critical to empowering artificial intelligence systems to handle complex tasks like driving safely down a city street.
The layoffs were first reported by Bloomberg Tuesday and confirmed by CNN Business.
Data annotation specialists use software tools to manually label objects in video clips collected from Tesla vehicles. The specialists label common road objects like lane lines, stop signs, traffic cones, curbs and traffic signals. The labeled data is fed into the artificial intelligence system so that it learns to accurately perceive its environment. The more correctly labeled data an AI system has, the better it can be.
Tesla has developed an automated way to do some of this labeling work in recent years, which allows the automaker to streamline its labor force.
“Without the auto labeling I think we would not be able to solve the self-driving problem,” Tesla CEO Elon Musk said at the company’s AI Day in August 2021.
Tesla executives have suggested that automating data labeling has already accelerated its work on self-driving vehicles.
Ashok Elluswamy, director of Autopilot software, said at the AI Day event that Tesla was able to collect 10,000 video clips from its cars and automatically label them in a week. Clips are generally 45 to 60-second video segments, as well as related GPS and odometer data.
“This would have taken several months with humans labeling every single clip,” Elluswamy said. He also described plans for producing a million auto labeled clips to “really crush this problem.”
There’s no clear cut answer to whether manual, human labeling or automated data labeling is more accurate, according to Raj Rajkumar, a Carnegie Mellon University professor who studies autonomous vehicles. Companies like Tesla may keep some humans involved to catch the flaws of the automated labeling, he said.
“If you do the economics of fewer humans, it’s a financial win,” Rajkumar said.
About 5 years ago, tesla trusted an outside commercial enterprise to label its self-using records, but moved efforts in residence in latest years, andrej karpathy, who leads ai at tesla, said closing 12 months. Records labelers have worked in san mateo, california and buffalo, big apple. He described this as essential to improving the satisfactory of tesla’s information. The agency built a group of extra than 1,000 people, he said at ai day in 2021. The process cuts and auto labeling do no longer take away the need for human labor.
In fact, tesla keeps to publicly submit some task openings for records annotation employees. “for us it is tons extra becoming a story of, ‘how do humans and computers collaborate to surely create those vector area information sets?'” karpathy said at ai day. Karpathy stated that tesla wanted the automobile labeling to be extraordinarily correct, which may have impacted how fast tesla has became to it. Synthetic intelligence specialists say there may be much less need for human annotators in the future as strategies are evolved that don’t require the expensive paintings.
“the destiny of records annotation is much less of it,” pedro domingos, a pc technological know-how professor on the college of washington, advised cnn business.
He cited the example of ai structures for language that analyze from loads of text statistics. Adella petrescu, a former autopilot records annotation manager at tesla, posted on social media tuesday that she were laid off. Petrescu stated she turned into promoted twice in one year and by no means had performance issues.
“for the past [two] years i’ve worked 50 [plus] hours every week, many weekends, and too many weeks with 16 hours days and all in favour of one cause that i sincerely believed in, i nonetheless do— [a] higher destiny for our subsequent generations,” petrescu said. Tesla did not straight away reply to a request for remark and commonly does not interact with the professional information media.

Related Posts

Leave a Reply

Your email address will not be published.