For just over 200,000 RMB, you can now own the most advanced AI-enabled electric vehicle on the market - the Geely Geometry 01. As stated by Lei Jun recently, "The past decade has seen the resolution of the electrification challenge in automobiles; the next decade will be about intelligence. Intelligence will be the decisive factor of this era." Currently, the Chinese new energy vehicle market is in a fierce elimination phase, where intelligence not only forms the basis of product competitiveness but also serves as the winning strategy for companies, ultimately determining the industry's dominant position in the long run.
What makes intelligence superior? Lei Jun's answer lies in "smart driving + intelligent cabin + ecosystem." And how can these aspects be strengthened? The Geely Geometry, backed by Baidu and Geely, demonstrates that true intelligence in cars can only be achieved through AI driving.
On March 25th, JueYue hosted the industry's first automotive AI technology conference, AIDAY2024. During the event, JueYue showcased how it deeply empowers its automotive robots with the cutting-edge AI native technologies from Baidu, bringing about a completely new and revolutionary evolution. Simultaneously, they announced the launch of the V1.4.0 new OTA version upgrade, offering over 200 functional optimizations and upgrades, once again demonstrating top-tier intelligent user experiences.
Jidrive Unveils Nationwide
Previously, with the support of Baidu Apollo's full-stack technology, known as the "West Point of the autonomous driving industry," the Jidrive 01 was launched online, delivering China's first BEV+Transformer pure visual advanced autonomous driving solution.
Soon after, Jidrive achieved another milestone by implementing China's first and the world's second mass-produced OCC (Occupancy Grid Map) pure visual system, completely integrating the algorithmic loop, bringing users integrated parking, highway, and urban full-scenario point-to-point navigation assistance.
During this AIDAY event, Baidu Apollo officially released the Visual Transformer Architecture (VTA) for autonomous driving, greatly enhancing the capabilities of vehicles in pure visual recognition, temporal tracking, real-time mapping, scene understanding, and more.
For example, the OCC visual occupancy grid map has been further upgraded, allowing automatic adjustment of resolution and frame rate based on different scenarios such as urban, highway, and parking lots, resulting in higher precision, more accurate models, increased obstacle recognition, and more flexible avoidance.
Meanwhile, leveraging large-scale models, Baidu has developed the industry-leading smart driving data production line and LLM-powered automatic driving data indexing. With generative AI technology, Baidu also possesses the capability to efficiently process long-tail data, which serves as vital data engines driving the development of end-to-end autonomous driving technology.
The Baidu Map LD map, generated by the large-scale models of automatic driving visual perception, is the native map for autonomous driving. It not only completes end-to-end rapid and scalable production but also covers over 360 cities as of the latest update, with updates at a rate of one city per day. Additionally, it incorporates key elements such as experience layers, safety layers, real-time layers, which are not present in high-precision maps, thus enhancing the intelligence and safety of autonomous driving.
The LD map boasts three key features: nationwide coverage, compatibility with both smart cabins and smartphones, and the safety ensured by having maps for smart driving. Thanks to this, wherever Baidu Maps can navigate, the advanced smart driving system of JiYue PPA can be activated instantly, providing synchronized 3D immersive lane-level navigation.
While competitors struggle, the SIMO voice assistant of JiYue 01 set a new industry benchmark upon its launch. It introduced the first-ever full-domain offline voice recognition, ensuring quick identification and response to in-car and out-of-car voices even without internet connectivity.
Now, Baidu's Chief Architect of Voice, Jia Lei, has officially unveiled a large-scale model for in-car voice recognition, becoming the first in the industry to tackle the challenge of building large voice recognition models. Next, a multimodal interaction solution integrating visual perception and voice will be implemented, leveraging the all-in-one in-car large-scale model to make the SIMO voice assistant smarter and more responsive.
In the upcoming V1.4.0 release, SIMO will fully integrate bidirectionally with Baidu Wenxin Yiyuan. It will support proactive recommendations, knowledge Q&A, car introductions, video searches, text-to-image generation, and one-click sending to mobile phones based on large-scale models, becoming the "encyclopedic steward."
It's worth mentioning that Jia Lei stated: "Even if competitors use the same Qualcomm 8295 cockpit chip, they cannot achieve the same SIMO voice assistant experience as Jiyue."
The reason is that from the outset of Jiyue 01's design, it has built a large-scale hardware infrastructure on the terminal side, "moving" the entire voice interaction system into the NPU computing unit. It integrates acoustic models, confidence levels, language models, semantic parsing, etc., into integrated modeling, and outputs results in parallel in one go, thereby achieving the ultimate interaction response speed.
Finally, Xpeng also announced the launch of its AI-based "Xpeng In-Car Home" ecosystem. By integrating seamlessly with Baidu's Xiaodu system, users can conveniently check their car's status, remotely control it, and receive alerts for any anomalies at home through Xiaodu speakers. This allows users to stay informed about their vehicle's real-time status with ease.
With the latest V1.4.0 version of Jueyue, which has fully activated push notifications, over 200 practical features have been upgraded, covering various fields such as intelligent driving, smart cabin, connectivity, and power systems, addressing each pain point raised by users.
For instance, there's a new OCC3D map menu, PPA urban navigation now covers all roads within the Fifth Ring Road in Guangzhou and Beijing; automatic turn signal function has been added; PPA highway navigation now supports customizable lane-change styles; turn automatic image assistance includes blind spot warning; the cabin introduces the "All-Weather Extreme Enjoyment" mode, suitable for camping, etc.; new features like in-car WiFi hotspot, wireless projection, SIMO fusion large model, Baidu Netdisk, and more.
By AI-refactoring intelligent driving and smart cabin features, and building an AI-driven IoT ecosystem, Jueyue not only fulfills its mission and original intention since its inception but also embodies the essence of rapid iteration and evolution of Jueyue's automotive robots. The strongest AI-powered car? It's none other than Jueyue.
