EqualOcean has learned that recently, the multimodal model company ChatImg(Chinese: 智子引擎) has completed the angel + round of financing.
ChatImg was founded in 2021, ChatImg is an early research team on multimodal models in China, with rich experience in large-scale model research, has insisted on continuous innovation for many years, launched a number of influential models, and maintained a leading position in the field of multimodal retrieval and multimodal generation. Prof. Lu Zhiwu of the Gaoling School of Artificial Intelligence, Renmin University serves as the company's chief technical advisor and commercially transforms large-scale model-related research results.
Compared with the large language model, which has begun to explore the path of success, the multimodal model is still almost blue sea. Based on more than three years of academic research, ChatImg team's multimodal exploration has begun to bear fruit:
In March 2021, the multimodal model "Wenlan BriVL (Bridging-Vision-and-Language)" was launched and iterated to version 2.0; on March 8, 2023, the multimodal conversation application with 15 billion parameters will be released. "Yuanchengxiang ChatImg"(Chinese: 元乘象) (prior to GPT-4) and the subsequent release of version 2.0 surpassed the current mainstream open source model in terms of Chinese-English bilingual multimodal dialog capabilities during the public evaluation of the multimodal dialog dataset (LLaVa).
However, stepping on the wind in the big language model, the progress bar of the multimodal model is also rapidly pulled from R&D to the landing stage. At present, To B is the main commercialization direction of ChatImg. Previously, ChatImg also reached cooperation with "iSoftStone Wisdom", an urban governance solution provider, to jointly develop an urban multimodal model and urban governance cloud.
ChatImg CEO Gao Yizhao said that in order to accelerate the landing of the multimodal model in the To B scene, this round of financing funds will mainly be used to expand the team and purchase computing power. "At present, in addition to computing power, the company's biggest cost is labor." However, Gao Yizhao believes that a complete product, business and engineering team is essential for the landing of applications in To B scenarios, "The development of products not only includes model R&D, but also engineering such as front-end page development, back-end concurrent maintenance, and so on."
This round of financing was jointly led by two leading investment institutions, followed by a number of parties such as SoftCompass High-Tech(Chinese: 软通高科), and Jinghe Capital(Chinese: 竟合资本) acted as the exclusive financial advisor. The proceeds of the financing will be used to further deepen the research of the multimodal model, expand the team, and purchase computing power.