Two years ago, Xiaodi Hou was forced out as CEO of TuSimple, a self-driving truck startup she co-founded. TuSimple later closed its U.S. operations and delisted from the stock exchange. AI animation and games in China.
Now Hou is back with Bot Auto, a new self-driving truck startup based in Houston. Bot Auto recently made its public debut with $20 million in pre-Series A funding. In a way, Hou is bringing the AV band back together. Most of his 40-person team is made up of former TuSimple engineers who were laid off in a round of cuts that hit the U.S. team.
Although much of the mission and team are the same, Hou claims the underlying technical architecture is different. The founders say Bot Auto has entered the second wave of self-driving car technology. Here, AI advancements over the past few years have made it easier and cheaper to accelerate AV development.
“We basically developed a machine learning infrastructure with eight people in four months,” Hou told TechCrunch. “In my previous life, I would have probably needed more than 100 people and it would have taken a year to develop.”
AV 2.0 era
Hou is not the only founder talking about the AV 2.0 era, which is attracting renewed interest from investors despite the failure of many self-driving startups.
TechCrunch recently reported that autonomous delivery startup Nuro is transitioning to a more streamlined AI architecture. Waabi, a self-driving truck startup that raised $200 million this year, also introduced a more advanced AI model. And Wayve, which raised $1.05 billion this year, is developing a self-learning system.
One of the AI advancements that has helped AV companies in this second wave, including Bot Auto, is Translator, the same architecture behind large-scale language models like OpenAI’s ChatGPT. Hou said Transformers allow Bot Auto’s massive neural network to take data from a variety of sensors, such as cameras, radar and LiDAR, to create a “world model” of the vehicle’s surroundings, while simultaneously recognizing objects and traffic signals and planning a route. .
Neural networks, which Hou calls “all the bases” models, fall short of decision-making. This is a type of data-in-decision flow that occurs in an end-to-end model, which Hou says creates a black box effect. Instead, decisions are made through “principles-based systems,” which Hou says leads to “explainable AI.”
According to Hou, Bot Auto can also speed up the data collection and processing process through pre-training, which trains neural networks on large datasets before being fine-tuned for specific tasks or environments.
“Pre-training is the key to fast-track training and jump-start your deep learning efforts without having to accumulate data for years and label it all,” Hou said. He noted that improvements in supervised self-learning allow Bot Auto to pre-train models from unlabeled data without human feedback, significantly speeding up the process.
Hou said Bot Auto doesn’t have to think about how to tweak GPUs or do high-performance data storage because it can leverage open source tools that weren’t available when he was leading TuSimple.
For example, according to Hou, Bot Auto uses Kubernetes as its cloud operating system to automate almost all critical data-related tasks, such as large-scale simulations and mining road test data.
Focus on operations
According to Hou, doing the front-loading work in building AV infrastructure means Bot Auto can focus more on its core business of bringing AVs to market.
“Of course we are always trying to keep up with the best technology, but we are not here to show off our technology,” Hou said. “We are here to facilitate operations and make it happen. That’s why we operate from Houston.”
Texas has become a leader in self-driving trucks not just because the state’s clear skies make testing easier. Texas has some of the most profitable freight routes in the country, including the I-45 corridor between Houston and Dallas. Waabi and Kodiak Robotics are testing their trucks on this route every day.
Now, so does Bot Auto.
Hou said his startup has begun testing autonomous trucks with human safety operators in the front seat on public roads in Texas and has already secured two letters of intent from customers in the logistics and transportation industries. Hou said Bot Auto is on the right track to complete a hub-to-hub demo in 2025.
Despite the scalability of its base model, Bot Auto is cautious about expanding its field operations.
“We will restrain ourselves in a very cost-effective way,” Hou said, noting that investors in 2024 are only investing in real, sustainable technologies. “We operate a very small fleet. We will focus on a few lanes and only start to scale once we know how to reduce costs and operate efficiently.”
Not without difficulties
Despite the win, Bot Auto is likely to face some headwinds.
Investors may be excited about the idea of a new era in AVs, but they haven’t forgotten the trials of the past few years, including the closures of Embark Trucks and Waymo’s autonomous cargo businesses.
And then, of course, there’s TuSimple’s failure. The worst of all happened after Hou was ousted as leader.
TuSimple’s board ousted Hou after an internal investigation revealed some overlap between the company and Hou’s co-founder, Mo Chen, with Hydron, a Chinese startup he founded to produce self-driving hydrogen trucks in 2021, around the time TuSimple went public in the United States. I decided to do it.
Hou has claimed he was fired without cause, but these public firings, and those fueled by geopolitical tensions, may give investors pause. This is a claim company sources previously backed up to TechCrunch. According to Hou, he has also learned from experience not to try to cross the line into unfriendly countries, which he said is one of the reasons why Bot Auto is a US-based company with no ties to China.
When asked why Hou hasn’t launched Bot Auto in China, where regulations for AV companies are more favorable, he said the higher labor costs for truck drivers in the U.S. increase the potential value of autonomous driving.
“It was a business decision,” he said.
Then there is the threat of legal action from TuSimple itself. After being fired, Hou retained his seat on TuSimple’s board, but resigned in March 2023 after the board accused him of trying to poach employees for his new venture. Hou denied this at the time and told TechCrunch that he disclosed his decision to start a new company after leaving the board. He said the TuSimple graduates who make up his staff today came to him because of their loyalty and desire to continue working in the AV field after being laid off.
It is unclear whether these accusations will resurface. TuSimple CEO Cheng Lu told TechCrunch that the company would likely sue Hou for IP and trade secret theft.
Mr. Heo said he was not worried.
“Everyone here is suing everyone,” he said. “And after accounting for the technology we’re building, most of the things we’re doing won’t even happen before 2021.”
