DeepSeek: A New Era of Smart Mobility

As the post-Spring Festival workforce resumes, a buzz surrounds the latest trending topic within the realms of traditional laborers and entrepreneurs, as well as both large corporations and small businesses—DeepSeekThe significance of DeepSeek is underscored by the fact that not only tech behemoths like Alibaba, Tencent, and Baidu have begun integrating the platform, but even automotive companies, seemingly unrelated to generative models, are vying to catch this wave of opportunity.

Reports suggest that over 20 mainstream automotive companies, such as BYD, Geely, Chery, Chang'an, and Dongfeng, have announced their intention to adopt DeepSeek's large modelThis trend indicates a transformative shift in the competitive landscape of China’s automobile industry leading up to 2025. With a price war cornering many traditional car manufacturers, there’s a surging interest among these companies to leverage artificial intelligence as a fulcrum for what they hope will spark a “second revolution” in smart driving technologies.

However, a pertinent question arises: can AI truly serve as a panacea to the intense competition and downward pressures within the automotive space? While the frenzy surrounding DeepSeek has turned it into a darling of the auto sector, the initial enthusiasm also raises skepticismThe auto industry is rushing to embrace DeepSeek in hopes of capitalizing on the ensuing traffic boom that comes with such technological integration.

In the current climate, topics associated with DeepSeek seem destined to trend, not just in automotive circles but across sectors like finance, home appliances, energy, telecommunications, and moreOver 200 prominent companies have stated their intentions to integrate DeepSeek into their operationsFor established firms, adopting DeepSeek doesn’t pose significant challenges; companies equipped with computational resources can easily accommodate demands well beyond what DeepSeek itself can handle

Advertisements

Furthermore, DeepSeek, being free and open-source, presents a compelling case for adoption—who wouldn’t want to join forces to potentially capitalize on its advantages without incurring costs?

On the flip side, automotive manufacturers aim to utilize DeepSeek to refine their own smart interaction systems, with a pronounced focus on enhancing user experiences through intelligent cockpit integrationThis pursuit appears to highlight a priority among companies such as Geely, Chery, Zhiji, Leap Motor, Great Wall, and DongfengTheir promotional efforts post-DeepSeek integration are heavily focused on the realm of smart cabins, with the ambition to achieve more nuanced and fluid voice-based interactions.

For instance, Geely has noted that the integration of their proprietary Xingrui model with DeepSeek-R1 has resulted in a 40% increase in response speed and an impressive 98% accuracy in intent recognitionSimilarly, Dongfeng has expressed that integrating with DeepSeek will make voice interactions within vehicles more intuitive, enhance situational understanding, and significantly speed up function iterations.

In recent years, many manufacturers have been racing to develop their own large models; however, the specifics of typical automotive scenarios tend to limit the application of these large modelsMost users generally approach such systems with straightforward inquiries about the weather or traffic conditions, leading many automakers to focus their R&D efforts primarily on advancing smart driving technologies while viewing complex interactive capabilities of generative models as merely sufficient for basic needs.

Therefore, while large models like DeepSeek have been "introduced into vehicles," the reality is often not much different from that of first-generation smart speakers, which typically function on a simple question-and-answer basis, lacking the full-fledged capabilities of an intelligent assistantHowever, with DeepSeek’s pronounced logic and reasoning abilities, there is hope that these automotive generative models could evolve to support multi-turn dialogues and contextual understanding, thus creating a more seamless user interaction experience.

Moreover, the open-source nature of DeepSeek serves as a pivotal advantage, enabling manufacturers to significantly dilute operational costs

Advertisements

For example, Geely has reported that by leveraging the dual-brain coordination of their Xingrui model alongside DeepSeek-R1, they can reduce expenses to just one-third that of traditional solutions.

Yet, the fact that many automotive companies are prioritizing DeepSeek for voice control and intelligent cabin training, while beneficial in enhancing user experiences, begs the question of differentiationCurrently, the concern remains that many could devolve into mere “question-answer machines.” Thus, amid the excitement, it is paramount to ask if DeepSeek can facilitate “AI democratization,” ultimately accelerating an equitable transition to smart driving and recalibrating the competitive landscape of the automotive industry.

In reality, there are not too many automotive companies actively deploying large models for smart driving just yet, with only a handful, including newer entrants like Li Auto and Xpeng, as well as tech giants like Huawei and Xiaomi, participatingMeanwhile, major traditional groups are still grappling with technological developmentHowever, since last year, following Tesla's announcement of the first end-to-end smart driving large model, new entrants have swiftly followed suit, shifting the entire narrative of domestic smart driving technology towards software-driven approaches.

Theoretically, this sort of end-to-end technology is ostensibly more cost-effective than perception models because it eliminates hardware costs, such as those associated with LiDAR, and reduces the reliance on manual coding, thereby lowering pre-training expensesNonetheless, the threshold for these costs remains significantAs Xpeng’s chairman He Xiaopeng stated, the end-to-end model raises the barrier for smart driving due to the need for increased funding, computational resources, and data.

This explains why even though "smart driving" has almost become standard across new-energy vehicles, the adoption rates for “high-tier smart driving” remain slow

Advertisements

This proposition gained considerable attention when BYD announced that its advanced “Heavenly Eye” smart driving assistance system would become a standard feature in cars over 100,000 YuanPreviously, Xpeng and Huawei had also been committed to pushing the boundaries of “smart driving equity,” with Xpeng lowering the bar to the 150,000 Yuan bracket and Huawei bringing its high-tier smart driving systems to vehicles priced at 200,000 Yuan.

However, the emergence of DeepSeek has prompted renewed speculation on whether smart driving equity can be further democratizedPresently, the consensus in the industry appears to be that large models are unlikely to quickly elevate the state of smart driving technologiesA report from Guolian Securities highlighted that DeepSeek could create high-fidelity virtual driving scenarios and utilize synthetic data to train intelligent driving models, thereby supplementing the lack of extreme scenario data in real-world road testing and enhancing model adaptiveness to complex environments.

Nonetheless, an advisory from Cui Dongshu, Secretary General of the Passenger Car Market Information Joint Conference under the China Automobile Circulation Association, cautioned that DeepSeek is not a panaceaIndustry insiders acknowledge that while applying DeepSeek's large model to smart driving is theoretically feasible, safety remains paramount and DeepSeek's response to high concurrency scenarios lacks the stability observed in traditional models.

As such, the likelihood of automotive companies fully integrating DeepSeek into smart driving systems remains slimNevertheless, the advent of DeepSeek has stirred up new ideas in the automotive sector, shifting from the early focus on hardware accumulation to a more people-centric approach—transitioning from competing solely on sensor hardware and chip capabilities to now striving for AI-driven advancements in smart driving.

Historically, the initial impetus for automotive tech progression evolved out of a focus on hardware accumulation and accurate data collection, which necessitated substantial investment, resulting in a high industry barrier to entry

However, in light of the “AI-driven era” on the horizon, competition based solely on funding and scale is steadily being challenged.

This signifies that future advancements in smart driving might not require “grandeur by titanic funding,” giving rise to smaller second-tier manufacturers capable of leveraging AI technologies to gain footholds in smart driving and fundamentally alter the competitive dynamics of the market.

While at this moment AI may not suffice to constitute a “cure” for smart driving challenges, it offers a glimmer of hope to allow more companies to take part in the race for smart technologies, potentially paving a way for genuine “smart driving equity” to take rootHowever, juxtaposed with the fervor shown by some mainstream manufacturers, newer entrants like NIO, Li Auto, and Xpeng, alongside tech giants like Huawei and Xiaomi, have chosen to adopt a “cooling approach,” indicative of an underlying tussle for ecological dominance among the industry’s players.

As previously mentioned, DeepSeek's large models have yet to find direct application within smart driving trainingFor mainstream car manufacturers, developing a connection to DeepSeek serves more to ameliorate their intelligent interaction capabilitiesHowever, companies like Huawei, Xiaomi, and NIO, which have developed multi-platform ecosystems, are reluctant to yield data to third-party models, as doing so would be akin to surrendering their organizational cognitive frameworks to another entity.

The high completion level of DeepSeek also creates potential risks for companies that become overly reliant on such technology, possibly causing stagnation in their internal development capabilities and possibly forfeiting technical controlA similar narrative has unfolded in the smartphone industry, where free open-source platforms parallel to DeepSeek have previously emerged, likened to the initial Android system, while companies focusing on their proprietary models and ecosystems mirror Apple’s IOS.

Though many smartphone manufacturers have leveraged Android to hasten their intelligentsia, the long-standing success of the high-end market resides only with Apple, the company that commanded the system and its affiliated chips

Advertisements

Advertisements

Your email address will not be published.Required fields are marked *