The Key Puzzle of AGI: The Exploration Path of Industrial 3D Generation and DeepDesign

In the just - passed "first year" of AI applications, the market's focus has shifted from general large - language models to real - world applications across various industries.

Among numerous AI startup sectors, the empowerment of the professional design field by AIGC has always been the area closest to commercialization, but its technological depth is also the greatest. Especially in the industrial 3D design field, most of the content generated by ordinary AI can only meet visual experiences and is rarely truly applicable to intelligent manufacturing.

Recently, DeepDesign, a domestic AI - powered industrial 3D design company, was selected as one of the TOP 10 projects in the second season of the "Lenovo New Business Innovation Ecosystem Roadshow". This is a strong proof of the value of AI applications in the industrial 3D design field and also reflects Lenovo's Small and Medium - sized Business Group's continuous exploration in the AI application field and its vision of accompanying customers' growth and success.

As the core platform for Lenovo Group to play the role of a leading enterprise in the industrial chain and drive the integrated innovation of startups, Lenovo's Star Program for Startups has established a three - dimensional empowerment system of technological support, industrial collaboration, and capital acceleration. Startups that join Lenovo's Star Program will have the opportunity to receive all - round and multi - dimensional support from Lenovo Group, including but not limited to co - creation in R & D, ecological openness, capital support, business expansion, brand and service empowerment, etc. Since the launch of the "Star Program for Startups", it has successfully helped hundreds of startups accelerate their development and has become an important link for Lenovo and small and medium - sized enterprises to advance hand in hand and empower each other. In the era of accelerating the cultivation of new productive forces, Lenovo's Star Program for Startups will continue to expand the recruitment scale of ecological partners and looks forward to building a value - sharing community with more industrial innovation forces and jointly climbing to new heights in global technological competition.

Reshape Industrial 3D Design with AI

Let's go back to 2016. Ma Jia, who had been deeply involved in the field of industrial digitalization since his postgraduate years and had been in this field for nearly a decade, started to implement his research results and practical experience in industrial 3D design efficiency with the help of his childhood friend (a double - doctor in computer science).

Industrial 3D involves 3D modeling for manufacturing and use in the physical world. The generated 3D objects need to have a complete internal structure to ensure functionality and manufacturability. Different from digital 3D generation, which only focuses on shape and appearance, industrial mechanical 3D design needs to consider the "internal organs" of the object to ensure its normal operation in the physical world.

As the understanding of deep learning becomes deeper, the application value of AI as a tool in industrial design has also emerged. However, most AIGC 3D projects in the market currently still focus on the digital 3D generation field. In contrast, the industrial 3D generation field is much quieter.

The reason behind this is that industrial 3D generation requires not only in - depth industry know - how but also the mastery of the most cutting - edge AI technologies. Both are indispensable.

"Before starting commercialization in 2022, our team had authorized and announced more than 30 invention patents. It was like spending several years getting a doctorate and then doing post - doctoral research," Ma Jia said straightforwardly. In those years when the team mainly focused on R & D and continuously iterated the product, in fact, no team in the market was clear about what technology would lead to the explosion of industrial 3D design in the future. Everything could only be learned, judged, and explored.

Ma Jia and his team firmly believe that the current position of DeepDesign is like ChatGPT on the verge of its launch.

Currently, DeepDesign consists of four subsystems, namely iCADS, DesignNet, ChatDesign, and DesignGPT. Among them, ChatDesign, or conversational design, uses the self - developed and pre - trained Colap - 3D model, allowing users to drive the design and modification of mechanical 3D through conversations; DesignGPT is a large - scale base model for mechanical 3D design in the true sense. The team has pioneered a new algorithm architecture globally, with a parameter scale of tens of billions. A preview version has been launched, and the official version will be released soon. It has received high recognition and affirmation from customers, investors, and the government.

"Before self - developing DesignGPT, we couldn't even find a relevant academic article. This shows that the technology and the market are very forward - looking. We are in an emerging field that has not been widely explored. As long as we give full play to our technological advantages, we have the opportunity to become pioneers in this field."

Brave the Unexplored Territory

As a startup whose technology has not been fully verified, in addition to dealing with loneliness and anxiety, it also needs to maintain a keen sense of the marketization trend of cutting - edge technologies. This is the reason why DeepDesign hopes to join the Lenovo ecosystem.

Although as a startup, its own funds and personnel are not abundant, Ma Jia found that the benefits of ecologicalization are very obvious: most of the enterprises contacted through the ecosystem have a strong awareness of AI application and will even take the initiative to discuss commercialization with DeepDesign.

There is still infinite room for imagination in the field of AI - powered industrial 3D design.

Data from the National Bureau of Statistics shows that in 2019, there were about 350,000 industrial enterprises above designated size (with an annual output value of over 20 million yuan) in China. Even if only 20% of these enterprises become the first to adopt AI - powered mechanical 3D design, the market space is already quite large. In particular, DeepDesign mainly focuses on pillar industries such as transportation equipment including aerospace, shipbuilding, automobile manufacturing, and rail transit, as well as mechanical equipment including power equipment and engineering machinery. The value of cost reduction and efficiency improvement brought by intelligent design is immeasurable.

DeepDesign has been in close contact with Lenovo Capital and Incubation Group. Whether from the technical or commercial side, the flexibility and stability of the Lenovo ecosystem have impressed Ma Jia deeply.

For example, after the emergence of ChatGPT, Lenovo responded quickly to the application of AI in the industrial field and started to build its self - developed infrastructure in a short period. At the same time, the Small and Medium - sized Business Division also began to gather the strength of a new group of "GPT entrepreneurs". As a leading enterprise in the front - end, Lenovo was able to recognize the importance of artificial intelligence earlier, which also increased the confidence of entrepreneurs like Ma Jia to continue.

Ma Jia recalled that when he first started to contact the Lenovo ecosystem, he deeply felt Lenovo's emphasis on and support for the innovation of small and medium - sized enterprises as a large enterprise with a wide - ranging layout in industrial manufacturing technology innovation. "Lenovo has a very open mindset. They are willing to use their ecological advantages to fully empower small and medium - sized enterprises with their accumulated capabilities and experience. This is very rare," Ma Jia added.

Ma Jia found that Lenovo's commercialization ecosystem has great ambitions. It does not only focus on the upstream and downstream of its own industrial chain but also "blooms in multiple points" in multiple cutting - edge technologies, especially in the field of artificial intelligence. This has subverted Ma Jia's brand impression of Lenovo and has become an opportunity for DeepDesign to participate in the second season of the "Lenovo New Business Innovation Ecosystem Roadshow" and find a way to collaborate with Lenovo through new connections.

"We are a team with technological ambitions. In addition to ambition, we must also be down - to - earth and have stable commercialization capabilities. For us, all three are indispensable. And in this roadshow, it is obvious that these three points are exactly Lenovo's requirements for its commercialization ecosystem and the assessment dimensions for entrepreneurs."

The Most Important Piece of the AGI Puzzle

The manufacturing rules and standards of the physical world are the territory that AGI must conquer in the future.

Currently, overseas industrial software giants such as Siemens, Dassault, and Autodesk are applying AI to CAD modeling. However, CAD and industrial software are bottleneck technologies for China, and the competition in intelligent CAD is not suitable for startup teams.

DeepDesign is exploring a brand - new path. Although this vision is still in the process of being realized, its potential impact may bring about far - reaching changes to the traditional CAD model, similar to how ChatGPT redefined search engines - it is both exciting and full of imagination. DesignGPT, which Ma Jia calls the "ultimate solution", can achieve efficient industrial 3D generation with limited data and resources, providing a new idea for solving complex industrial 3D generation problems.

In 2025, finding application scenarios is the most important thing for most AI entrepreneurs, but DeepDesign is different.

The problem this enterprise faces is that the application scenarios and requirements in the mechanical design field are very clear, and there is also a quite large commercialization space. However, in such a grand design scenario, achieving a passing score in the generation effect is often "meaningless".

Therefore, for an AI product to make an impact in the industrial software field, it must reach a score of 90, allowing designers to only worry about extremely minor modification possibilities, which can truly liberate productivity and imagination. In the process from 60 to 90 points, the charm and perseverance of entrepreneurship will be fully revealed.

For an ecological enterprise like Lenovo, building an AI ecosystem from a technological and commercial perspective is only the first step. The next step is to pave the way for these high - quality entrepreneurs who are "closer to the future" from the terminal to enterprise customer resources and provide new value connection points, which is also its responsibility in this era.