MOUNTAIN VIEW, Calif.—The autonomous driving software developer Ghost Autonomy shut down its worldwide operations in early April.
The company had about 100 employees and was based in Mountain View, California, with additional offices in Detroit, Dallas, and Sydney, Australia.
Ghost Autonomy’s goal was to provide the first self-driving technology for passenger cars that does not require your attention.
Founder and CEO John Hayes told The Epoch Times in an email, “Ghost is completely shut down, remaining funds will be returned to investors, employees are figuring out their own path (we were able to pay severance) and the IP is available for commercial and research purposes.”
Ghost Autonomy’s press release about the closure states: “The path to long-term profitability was uncertain given the current funding climate and long-term investment required for autonomy development and commercialization. We are exploring potential long-term destinations for our team’s innovations.”
The company was founded in 2017 by Mr. Hayes and Volkmar Uhlig, who over time received $220 million in total funding. Investors included Mike Speiser at Sutter Hill Ventures, Keith Rabois at Founders Fund, Vinod Khosla at Khosla Ventures, and the investment firm Coatue.
They also received a $5 million investment from the OpenAI Startup Fund in November 2023 to bring large-scale multi-modal large language models (MLLMs) to autonomous driving.
The company was building a software-based solution that combined artificial intelligence with high-volume mobile sensors and chips to make autonomy possible for consumer vehicles.
Ghost Autonomy stated that MLLMs have the potential power to reason, combining perception and planning to give autonomous vehicles a deeper understanding and guidance on the correct driving maneuver by considering the scene in totality, accepting image and video inputs alongside text.
The company stated that its large models have encoded a generalized “understanding of the world” and can understand the concept of a driving lane, while a single task model is only trained to find white painted road markers on the road.
Ghost Autonomy stated, “As a result, the large model will perform much better in ambiguous circumstances found in the real world, like when the paint is faded or splattered, or construction barrels have altered the path of travel.”
Mr. Hayes stated: “A lot of people and companies see a future for consumer autonomy, certainly the expanding operations of Waymo provide a glimpse into the future. The consumer challenge has a triple challenge: reliability acceptable to all stakeholders, widespread operations without remote monitoring, and low cost with low maintenance. Companies pursuing the space have typically chosen a subset of these problems to solve. I expect the auto companies will first solve the cost and widespread operation problems by the end of the decade (with driver monitoring) and then solve reliability through massive observation into the next decade.”
According to an analysis from Cybertalk, there are still concerns about autonomous vehicles. One concern is that the increase in autonomous vehicles could cause job displacement for truckers, delivery drivers, and taxi drivers.
Another concern is the uncertainty of what an autonomous driving system might do when faced with moral decisions in real time, such as whether to prioritize occupants or pedestrians during unavoidable accidents.
Also, when an autonomous vehicle is in an accident, determining who is at fault could be difficult.
With autonomous vehicles connected to the internet, they’re vulnerable to hackers and cyber attacks.
They’re also susceptible to software glitches and can become parked in the road due to complex street conditions, causing backups.
According to AAA’s latest survey on autonomous vehicles, 66 percent of U.S. drivers express fear of fully self-driving vehicles and 25 percent express uncertainty.
