Author | Li Dongmei
Ghost, a self-driving car software company, has announced its collapse and laid off hundreds of employees
Recently, according to a number of foreign media reports, an autonomous driving company invested by OpenAI has now closed.
Founded in 2017, the Mountain View, California-based Ghost Autonomy company is a startup developing self-driving software for automaker partners.
The company announced on its website last week that "Ghost Autonomy has shut down its global operations and shut down the company on April 3." We are proud of the substantial technological innovations and advancements that the Ghost team has made in our mission to realize the software-defined consumer. Given the current financing environment and the long-term investments required for in-house development and commercialization, the path to long-term profitability is uncertain. We are exploring potential long-term goals for our team's innovation. ”
The company's statement continued: "Thank you to the employees, investors, and partners who helped bring Ghost's vision to life. We really appreciate your support of Ghost along the way. ”
Founded by Hayes and Volkmar Uhlig, Ghost Autonomy has around 100 employees in its seven-year history and offices in Mountain View, Dallas and Sydney.
The company's co-founder and CEO, John Hayes, founded Pure Storage in 2009 and took the company public six years later.
Ghost 联合创始人兼 CEO Jay Ohn Hayes
As Pure's Chief Architect, he leveraged the consumer industry's transition to flash to reimagine data center storage and invented the ultra-fast flash storage solution, currently operated by the world's largest cloud and e-commerce providers, financial and healthcare organizations, scientific research organizations, and governments. Hayes wanted Ghost to redefine commercial consumer hardware by using software to achieve near-perfect reliability and redefine commercial consumer hardware, just like Pure.
Like many startups trying to commercialize self-driving car technology, Ghost has changed its business strategy over the years. The startup was originally called Ghost Locomotion and was founded in 2017. Two years later, the company made its public debut, receiving a total investment of $63.7 million from companies such as Rabois of the Founders Fund, Vinod Khosla of Khosla Ventures, and Speiser of Sutter Hill Ventures, as well as plans to develop a kit that would allow private passenger cars to drive themselves on highways. The company says it will offer the technology in 2020.
Ghost has been emphasizing that its goal is to provide technology for consumer cars purchased by regular customers. At the time, Jacqueline Glassman, Ghost's general counsel, explained, "Unlike robo-taxis or delivery services, Ghost offers self-driving services for cars that people own and drive to work every day. Offering its services in software form, Ghost is designed for the mass market, partnering with automakers to bring autonomous driving into the mainstream. ”
In addition, the company is keen to point out that their path to development is different from that of its competitors, and that they are focusing on artificial general intelligence rather than using image localization to identify potential obstacles to ensure that self-driving cars avoid them. Ghost is more focused on tracking clusters of pixels in future scenes — in short, it doesn't matter what the obstacle is, it's important that it's an obstacle and that you can avoid it.
But after that deadline reached in 2020, Ghost changed its initial strategy to focus on collision prevention technology instead, and raised another $100 million in 2021.
The Series D round was led by Sutter Hill Ventures, with participation from Founders Fund and Coatue. In an interview with TechCrunch media back in 2021, Hayes said that the startup is not completely closing the door to the consumer kit model, but is turning its attention to universal collision avoidance technology to get to market faster.
Founded 7 years ago and delivered 0 years, it has attracted attention for being invested in by Open AI
In fact, Ghost's lofty ideals and visions are not backed up by sufficiently advanced and innovative technology. Although it has raised 220 million US dollars (about 1.5 billion yuan) in the seven years since its establishment, this amount of money is a drop in the bucket in the extremely money-burning field of autonomous driving.
Last year, Hayes said in Forbes magazine, "Automotive companies have struggled with software development for the past decade, both in terms of execution and customer adoption. Consumers are very disappointed with the new technological features in their cars. Software is actually hard to develop for software companies – even the giants of Silicon Valley. ”
Such a company, which does not have much amazing in terms of technology or founder's background, is really getting attention because it has been invested in OpenAI, an AI superstar company.
Last November, just five months before the company shut down, the startup partnered with OpenAI through the OpenAI Startup Fund to gain early access to Microsoft's OpenAI systems and Azure resources. Ghost also received a $5 million investment from OpenAI. Last year, the company closed an initial funding round of $55 million from early-stage investors such as Keith Rabois of Founders Fund and Mike Speiser of Sutter Hill Ventures.
OpenAI and Ghost announced the move at the same time, with plans to introduce large language models (the technology behind ChatGPT) to autonomous driving.
At the time, Hayes touted the company's plans to explore the use of multimodal large language models (LLMs), artificial intelligence models that can understand text and images, in autonomous driving. He believes that large language models provide a new way to understand the "long tail", adding inference to complex scenarios where current models are insufficient.
Brad Lightcap, COO of OpenAI and manager of OpenAI's startup fund, also stood for the platform, saying: "Multimodal models have the potential to extend the applicability of large language models to many new use cases, including autonomous driving and automobiles." Multimodal models are able to comprehend and draw conclusions by combining video, images, and sound, so they may create a new way to understand scenes and navigate complex or unusual environments. ” 。
Prior to the collapse, Hayes also sent an email to TechCrunch saying the company had completed a highway driving product and was moving through urban environments with what he described as "last-mile delivery."
But ultimately, Hayes said, "for years, Ghost Autonomy has not been able to finance itself or achieve the level of engineering intensity needed to bring a product to the consumer car market."
That said, 7 years after its founding and burning $220 million, Ghost Autonomy still hasn't had a single product hit the market.
How far are we from Level 5 autonomous driving?
A lot of people want faster car development cycles, but the industry needs to get it right the first time – customers won't tolerate semi-finished products. Making one of the most complex industries work requires incredible cross-cutting engineering, and this can only be achieved in extraordinary time.
The development of self-driving cars is not going well. In addition to the news of Ghost's collapse, a recent example is Apple, one of Silicon Valley's biggest giants, laying off more than 600 employees in Santa Clara, according to state filings, with employees at Apple's eight offices in Santa Clara being laid off.
Last fall, General Motors' Cruise self-driving division came under regulatory scrutiny for one of its vehicles crashing into and dragging pedestrians, leading the U.S. automaker to lay off employees and cut $1 billion in spending. But Mary Barra, GM's chief executive, said in February that "the benefits of self-driving technology" were "enormous" and that the company had "extremely valuable assets."
According to a report by research firm GlobalData, drivers won't be able to experience the road to a fully autonomous vehicle for at least 20 years, and the road to Level 4 and above autonomous vehicles "is likely to be slow."
Level 5 autonomy involves self-driving cars that don't require any human interaction, which means that when eventually deployed, they will have no steering wheels or pedals.
"However, it is reasonable to assume that the new Level 4 vehicles introduced in 2035 will be significantly better than those introduced in 2025, so the path to Level 5 is likely to be progressive," the report states. ”
Original link:Founded 7 years and 0 delivery, the OpenAI-invested autonomous driving software company announced its collapse after burning 1.5 billion yuan_Autonomous driving_Li Dongmei_InfoQ Selected articles