H3C Group, an affiliate of Unigroup, has announced the launch of LinSeer, a large-scale customized AI model aimed at government and corporate clients. This unveiling occurred at the annual Navigators Summit on June 7, following the recent release of large-scale general models by various internet firms and AI companies.
H3C finds itself in competition with these large-scale general models in specific industries, while it also provides them with hardware products. On June 9, H3C rolled out a range of products related to large-scale models, which included AI servers equipped with Nvidia's most recent H800 GPUs and an 800G silicon photonic data center switch boasting high throughput and low latency.
Developed using H3C's proprietary technology, LinSeer is specially designed to cater to the needs of distinct industries or regions. It operates within a private domain where client data is stored, thereby enabling the model to handle unique queries in its specific field. However, the model's main drawback is its inability to perform general-purpose tasks.
H3C had previously publicized its collaboration with Baidu's large-scale model ERNIE on February 20, playing a significant role in its initial implementation.
According to Zhang Li, President of H3C Group's Industry Business Group, amidst the booming generative AI, H3C is committed to supporting the application of large-scale models in the internet industry. It offers computational support like GPUs, which have witnessed a 300% surge in demand over the previous year.
The company also plans to collaborate closely with leading large-scale model manufacturers, such as Baidu, to provide intelligent solutions across various industries.
During H3C's event on June 9, Shen Dou, President of Baidu's Intelligent Cloud Business Group, outlined three key factors for enterprises choosing base large models: overall performance, iteration speed, and the completeness of the toolchain.
Shen noted that due to the high costs and technical barriers involved in their development, only a few companies with ample computational power, data, and a reservoir of high-end AI talent would be capable of creating and sustaining large-scale models.