Waymo, a robotaxi company of Google, recently announced that it will expand its autonomous driving service area to the San Francisco Bay Area. This marks the beginning of robotaxi operations in Silicon Valley, a symbol of U.S. innovation technology and big tech. Saswat Panigrahi, product lead at Waymo, emphasized that "launching a fully autonomous ride-hailing service in Silicon Valley is an important milestone."
In China, major automobile manufacturers, including BYD, announced that they will equip their self-driving cars with software developed by DeepSeek. DeepSeek has shaken up the generative artificial intelligence (AI) market previously dominated by the United States, igniting competition for AI supremacy. This means that Chinese automobile manufacturers are preparing to enhance their self-driving car technology with the support of DeepSeek.
AI technology is emerging as a game changer that will alter the competitive landscape of self-driving car technology. Development of autonomous driving technology began over a decade ago, but it is accelerating now with the rise of generative AI in 2022.
Will Lin, head of automotive at Applied Intuition, met with ChosunBiz at a hotel in Pangyo, Gyeonggi Province, on the 13th and noted, "AI is helping developers make overall software-related tasks more efficient," adding, "By actively utilizing AI, cars will become smarter and their performance will improve."
Applied Intuition is the world's number one company in self-driving simulation. Established in 2017, it has yet to reach a decade of existence, but its corporate value has reached $6 billion (approximately 9 trillion won), and 18 of the top 20 global automobile manufacturers use Applied Intuition's software based on revenue.
Applied Intuition gained recognition thanks to AI technology. The company creates a simulation platform that accurately replicates real-world conditions, training self-driving software on it. Self-driving vehicles cannot train on actual roads due to safety issues. The completeness of the self-driving software depends on how well the simulation platform mimics reality, and Applied Intuition is ranked number one in this field.
Will Lin attributed the secret to AI. He stated, "Applied Intuition gets help from OpenAI CEO Sam Altman and actively employs AI experts and engineers, leveraging AI to a great extent," adding, "We are not just a software company but should be considered an AI-based software company because we create software that combines AI expertise."
The emergence of generative AI is greatly aiding the development of simulation platforms and boosting the fidelity to levels resembling reality. In these platforms, a key aspect is not just drawing maps but confirming how self-driving software responds when unexpected variables arise, mimicking real road scenarios. For instance, a pedestrian or a bicycle might suddenly appear from behind a parked vehicle or a car might stop abruptly instead of proceeding through a yellow signal.
In the past, humans had to create such variables manually one by one; now, with the introduction of generative AI, it is possible to add variables much more intricately and rapidly, even conducting driving simulations.
An American corporation named VERSES AI recently showcased technology that predicts and prevents accidents by anticipating pedestrians or bicycles suddenly emerging from behind parked vehicles with self-driving cars. Hari Thiruvengada, CTO of the company, explained, "This AI algorithm has added 'occlusion reasoning' capabilities to predict the behavior of hidden road users, meaning it can even anticipate the movements of cyclists hidden behind parked trucks."
'Occlusion' reasoning poses significant challenges for AI models. This is due to the large amount of computation required and the likelihood of encountering untrained scenarios. VERSES AI tackled this issue using generative AI technology, allowing self-driving cars to autonomously adjust their speed and driving direction when approaching areas where invisible people or objects are likely present. Simulations utilizing Waymo's self-driving car data showed a high probability of predicting situations where pedestrians suddenly appear at walking speeds.
Utilizing generative AI is also a hot topic in the domestic self-driving industry. However, without indigenous generative AI like GPT or DeepSeek, South Korea inevitably faces a handicap compared to the United States and China. Experts emphasize that active government support and regulatory improvements are necessary to overcome this handicap. Yu Si-bok, a researcher at the Korea Automotive Technology Institute, pointed out that government projects related to autonomous driving were conducted before generative AI gained momentum, making additional investments challenging.
Shin Jun-woo, head researcher at Seoul Robotics, stated, "We plan to introduce LLM for additional safety in the long term, but there are practical limitations such as workforce shortages and lack of clear regulations on responsibility regarding the operation of autonomous driving systems," adding, "The introduction of a driving operator system responsible for real-time monitoring of self-driving vehicles and managing driving infrastructure, optimization, and verification is necessary."
Will Lin emphasized, "Although there are no indigenous generative AI models in Korea, there are various models available for use as open source, and the capabilities of AI experts are adequate," adding, "Because the autonomous driving industry is a complexly interconnected ecosystem, it is important for Korea to identify necessary technologies and evolve together with many partners."