Article source: i dark horse (ID: iheima)
On September 5, 32 AI-Native entrepreneurial projects launched a more than 4-hour "Wenxin Cup" roadshow at Beijing International Trade Center.
Peng Kangwei, born in 1998, is introducing her entrepreneurial project-Genie AI. "Generative AI can not only help people draw and make animations in the second dimension, but also help people expand three-dimensional IP and become personal movie directors."
As early as May 31, Robin Li said at the J.P. Morgan Global China Summit: "In the future, China will have its own ecosystem, with its own large language model and basic model. There will be at least one, and possibly two or two, in China. Three basic large models can support people in developing various AI native applications."
In the following more than three months, nearly 1,000 projects signed up for the competition, and finally 15 potential projects entered the "finals." ChatDOC , a professional knowledge Q&A assistant and reading platform based on document intelligence, and Buysmart.AI, an intelligent e-commerce shopping buyer based on large models, won the first prize, and the other teams won the second and third prizes respectively.
According to i Dark Horse’s observation, 80% of the contestants focus on large model applications, and the contestants are mainly those born in the 1990s. 2B and 2C applications each account for half of the country, and the industry distribution is diverse. It not only includes AI native applications such as intelligent assistants, community social networking, and productivity tools, but also covers the reconstruction and transformation of applications in medical, education, finance, legal and other industries. Half of the founders of the 15 projects that finally competed for the award were born in the 1990s.
Wu Hequan, an academician of the Chinese Academy of Engineering, once said: "The intelligent era is coming to us. It is more ambitious than the era 20 years ago. The next 10 years will be a critical period for the development of the intelligent economy."
At the "Wenxin Cup" Demo Day on September 5, i Dark Horse discovered many projects full of business prospects and innovative imagination. They may be the representatives and epitome of the intelligent era in the next 10 years...
01
Buysmart.AI：
Intelligent e-commerce shopping buyer based on large model
Buysmart.AI is positioned for smart e-commerce shopping buyers. Based on large models + self-developed recommendation technology, it provides users with a one-stop solution for shopping intention understanding, network-wide information collection, decision-making, and shopping guide.
"We are mainly engaged in overseas AI e-commerce shopping guides and smart e-commerce shopping buyers, which can assist users in selecting products and making shopping decisions. For example, users can tell the AI that I want a comfortable bicycle seat and a yellow one that looks good on me. T-shirt, it will recommend suitable products to you." founder Chen Lifei said.
Chen Lifei is also a post-90s generation. He graduated from Carnegie Mellon University. He has worked in Xiaohongshu, Meta and Pinterest and has extensive experience in the fields of advertising ranking, NLP and personalized recommendations.
In his view, the large language model is a tool in itself. "Our core lies not in the large language model itself, but in how to make AI capable of playing the role of a human shopping guide. The key is to ask the right questions and launch the right products."
Buysmart.AI company's products can conduct multiple rounds of dialogue with users, and match users with appropriate products and purchase links based on crawled e-commerce data and self-developed recommendation algorithms. Its products can better compensate for the large model in search and recommendation shortcomings.
Chen Lifei gave an example. "For example, I now have a budget of 10,000 yuan and let AI help me buy the most suitable motorcycle. We hope that the smart buyer product can crawl and learn relevant data and knowledge from the entire Internet. Become a professional buyer and finally tell the user that I have looked at all 100,000 motorcycles on the market and this one is the most suitable for you. When can we say this, when will our task be completed? finished."
Buysmart.AI used GPT before, and Baidu’s Wenxinyiyan was their first attempt to use a large domestic language model. Chen Lifei said: "It is not as difficult as expected to replace an overseas model with a domestic model. After we changed it, we didn't need to make too many changes, and the whole process was able to run."
02
Nolibox：
High-quality AI graphic design and creative generation platform
At previous industry summits, i Dark Horse listened to Nolibox’s sharing. Nolibox has also received angel round investment from Chuxin Capital and Pre-A round investment from Hillhouse Ventures.
The founding team of Nolibox comes from Tsinghua University and Peking University. CEO Xu Zuobiao also participated in this "Wenxin Cup" competition and road show.
It is understood that Nolibox Computational Aesthetics is committed to using artificial intelligence to empower design creativity. Its two products, "Hua Universe" and "Picture Universe", provide full-link AIGC and artificial intelligence design services to individual users and corporate users.
"We lived in a two-bedroom apartment in Huaqing Home and began to explore and develop the integration of artificial intelligence and creativity." Xu Zuobiao said: "The products we entered this time are to empower the creative scenarios of e-commerce and marketing. A super production tool. It has two main highlights, one is high quality, and the other is universal benefit."
In Nolibox’s view, the e-commerce marketing scene is a large and long-term valuable track. AIGC allows production materials to go from limited to unlimited in the past, and from high cost to low cost. At least 90% of the commercial value has not yet been tapped.
Will large models in the future subvert the domestic design industry?
Xu Zuobiao believes that it cannot be said to be subversive, but more of a boost or to unlock design and creative capabilities. AI is more like a tool. The essence of a tool is to help humans rather than subvert or eliminate humans.
"Actually, graphic design and creative drawings are two different things. The Stable Diffusion that everyone is familiar with now focuses more on creative drawings and has basically nothing to do with graphic design. It cannot be said that it has subverted designers or eliminated designers. More It’s just laymen watching the excitement.” Xu Zuobiao added: “Of course, it is also very dangerous if designers never embrace AIGC. Just like in a war, if you have to use cold weapons to fight hot weapons or nuclear weapons, it’s okay. , but the ending could be tragic.”
03
Magic technology:
AI for science artificial intelligence new material research and development
Not long ago, the room temperature superconducting material LK-99 proposed by South Korea attracted widespread attention.
Huanliang Technology also used LK-99 as an example in Baidu's competition.
There is a lot of trial and error and random processes behind traditional material exploration. The advantage of AI is that it is mainly data-driven. By analyzing past experimental data and literature and finding connections, it can help quickly screen materials. But the field of room temperature superconductivity itself is the Holy Grail of physics. Data limitations will greatly limit the application of machine learning. If it is applied through pure AI, there is a lack of understanding of physics or underlying mechanisms.
In addition to the two basic paradigms of experimental science and theoretical science that have been mainly used in scientific research in the past hundreds of years, now we may usher in a new paradigm: AI for Science - artificial intelligence will become a new production tool for scientists after computers.
Founded in 2021, Huanliang Technology is a data-driven material information design and engineering technology company.
According to reports, Huanliang Technology is based on materials informatics, combined with AI, computational physics, high-throughput experiments, etc., and integrates cross-field data to accelerate the industrial implementation of new material research and development in various industries. The company has completed POC for more than 5 companies, saving customers R&D costs while ensuring the improvement of material performance.
The team of Huanliang Technology has assembled many star members in the field of AI materials informatics. The company's founder and CEO Liu Yuyang once served as a postdoctoral researcher in the Department of Materials Science at MIT. He holds a PhD in theoretical physics from the University of Cambridge. His research has been cited more than 800 times in the fields of physics, materials, and biology. Team members mainly come from world-renowned universities such as MIT in the United States and Cambridge in the United Kingdom.
Huanliang Technology completed tens of millions of yuan in seed round financing in 2022, with investors including Sequoia China Seed Fund, Nest Capital, Zhen Fund and NewLight.
04
Conclusion
Shen Nanpeng once said that from steam engines to the Internet, every technological revolution is marked by the application and industrialization of new technologies, and digital intelligence represented by AI technology is no exception.
Through the Baidu "Wenxin Cup" competition and roadshow, we can see that there are chatDOC (tools), Nolibox (design assistant), Buysmart.AI (e-commerce shopping guide assistant), etc., which are full of business prospects, as well as expanding industries Imaginative Gennie AI (two-dimensional community), Magic Technology (AI+new materials), Null Hypothesis (AI+medicine), etc.
Startup projects represented by AI Native are rapidly emerging in China. A new wave of artificial intelligence entrepreneurship and innovation is sweeping across thousands of industries and is being implemented in various industries and scenarios.
Robin Li believes that AI native applications should meet three basic conditions: first, they can interact with natural language; second, they make full use of the understanding, generation, reasoning, memory and other capabilities of large language models; third, the interaction of each application does not exceed two levels. menu, so the interaction will be simpler.
From the basic layer to the model layer to the application layer, China's generative AI industry chain has been formed.
Compared with the large-scale model of "Wen Xin Yi Yan" released by Baidu, this "Wen Xin Cup" has demonstrated strong ecological influence. Empowering thousands of industries and enabling AI Native entrepreneurs, Baidu is accelerating the construction of a large AI model ecosystem.
It is understood that Baidu has provided the Wenxin large model API and Baidu intelligent cloud computing resources that can be stably called to the finalists. The winning project will receive an investment of up to tens of millions from Baidu. At the same time, Baidu will invest in technology, products, and development. Long-term support in terms of strategy and capital cooperation.
Robin Li once pointed out, "I am very optimistic about the development prospects of AI applications in China. If we look back on the history of the past few decades, everyone in China is very willing to embrace emerging technologies. Although we did not invent Android, iOS or Windows systems, we have developed many very Innovative applications, such as WeChat, Douyin, Didi, etc., and many other applications are very popular and practical. In the era of artificial intelligence, the same situation is true. Technology has brought many possibilities, and we are very good at developing applications and Take full advantage of these possibilities."
We can look forward to thousands of AI-native applications and new “Douyin” and “WeChat” emerging soon.
Click "Read the original text" to cooperate and consult immediately