Multiple car companies have announced the integration of AI functions, intensifying the competition in the already cutthroat auto industry. The trend towards intelligent vehicles is inevitable, and the ultimate goal of upgrading the next generation of computing architecture is to transform cars from cold machines into versatile assistants that can think, interact, and provide warmth.
After ChatGPT became popular, the journey of AI's subversion of the world is visibly accelerating. A Wall Street Journal article previously cited a McKinsey report pointing out that AI-related applications will bring annual growth of $2.6 trillion to $4.4 trillion to the global economy, equivalent to the annual growth of the UK's GDP.
For car companies that make intelligentization their main selling point, they naturally cannot fall behind in the new round of technological change. Recently, many car companies have announced the integration of AI functions, and the internal competition in the car industry is intensifying.
GPT Lands in Many Car Companies
Since the beginning of the year, with ChatGPT's popularity across the Internet, many car companies have begun to promote large models.
This Saturday, Li Auto held its first Family Technology Day at the LI AUTO-W AI Dream Factory (LI AUTO-W Changzhou Factory). In addition to announcing the first pure electric car model LI AUTO-W MEGA, LI AUTO-W also demonstrated several upgrades of its in-car AI assistant "LI AUTO-W Classmate" at the Family Technology Day. Now, LI AUTO-W Classmate has multi-modal perception capabilities, not only can execute common instructions such as "open this" and "open that" proposed by the car owner, but also carries its self-developed "Mind GPT" cognitive large model. Passengers can use dialects such as Cantonese, Shanghai dialect, and Sichuan dialect to ask LI AUTO-W Classmate various questions, and even tell stories and draw pictures.
In addition, in May, according to multiple media reports, NIO-SW has applied for trademarks such as "KnowmeGPT", "NIOGPT", and "NOMIGPT", involving social services, building repairs, scientific instruments, and other fields. The outside world judges that NIO-SW may also be planning to integrate large models into the vehicle system.
Earlier in March, Jidu Auto also announced that it will integrate the comprehensive capabilities of Baidu Wenxin Yiyu into a large model artificial intelligence interaction experience tailored for intelligent automobile scenarios.
In addition to new forces, the "old forces" of car manufacturing are also catching up.
Just last Thursday, Mercedes-Benz and Microsoft announced that they will add GPT functions to Mercedes-Benz cars in the United States. With the help of OpenAI and Microsoft's technology, Mercedes-Benz's original in-car voice assistant will be upgraded.
Previously, Mercedes-Benz owners could use the "Hey, Mercedes" command to control the voice assistant to complete some simple tasks, such as adjusting the temperature inside the car, finding a navigation destination, or making a phone call. Microsoft said that ChatGPT will make these voice commands smoother and more natural, and can achieve more functions.
With the support of GPT, the Mercedes-Benz voice assistant can remember the content being discussed and engage in multiple rounds of conversation with the driver or passengers.
Previously, General Motors also stated that it is working with Microsoft to explore the integration of ChatGPT in cars.
The Future is Here: AI+ Hardware Will Become a New Species
From the current perspective, the main reasons for car companies to integrate large models are: 1) to optimize intelligent driving, and 2) to create AI virtual assistants.
The GPT model mainly comes from the Transformer model. Transformer has achieved excellent results in the field of computer vision. For example, Tesla's BEV technology.
BEV can convert 2D images taken by cameras into 3D images, and unify them to a bird's-eye view for processing, forming a "God's perspective". Based on this, computers can learn to judge the driving environment in the 3D world like humans, and ultimately drive like humans.
The NOA navigation assistance driving system is also based on BEV perception and Transformer model.
In addition to autonomous driving, perhaps the more critical aspect is to create AI virtual assistants.
Although at present, the GPT storytelling and drawing functions demonstrated by car companies such as LI AUTO-W have more gimmicks, the combination of AI+ hardware is undoubtedly moving in the right direction.
This is also the consensus of car companies. Intelligence is the trend, and the endpoint of the upgrade of the new generation of computing architecture is to make cars no longer cold machines, but become a versatile assistant that can think, interact, and have temperature, a machine with wisdom, a new species.
LI AUTO-W and other car companies have already taken the first step towards this new species.