2024CTIS-文章详情页顶部

Zhipu Accelerating Deployment in Text-to-video Technology

(AsianFin)—Beijing Zhipu Huazhang Technology Co., Ltd. (Zhipu AI) completed a new round of financing earlier this year, with the Beijing Artificial Intelligence Industry Investment Fund as an investor.

Founded in 2019, Zhipu AI started from the Knowledge Engineering (KEG) Laboratory established in 1996 at Tsinghua University. The laboratory focuses on researching internet-related knowledge engineering and has published a series of leading research results in knowledge graphs, graph neural networks, and cognitive intelligence.

Zhipu AI is also one of the earliest companies in China to make inroads into the large model sector and is currently one of the leading large model manufacturers in China.

In March 2023, Zhipu AI launched the open-source foundational dialogue model ChatGLM series and completed four upgrades throughout the past year, and released the GLM4 series at the end of last year.

Leveraging the capabilities of fully self-developed foundational large models, Zhipu AI has also constructed an AIGC model product matrix, including the AI efficiency assistant Zhipu Qingyan, the high-efficiency code model CodeGeeX, the multimodal understanding model CogVLM, and the text-to-image model CogView, among others.

On Thursday, on the occasion of the one-year anniversary of ChatGLM's release, Zhipu AI CEO Zhang Peng said in Beijing that in 2023, Zhipu AI has had over 2,000 ecosystem partners and over 1,000 large-scale model applications, with over 200 companies engaging in deep collaboration with Zhipu AI, covering multiple leading enterprises in various sub-scenes such as media, consulting, consumption, finance, new energy, internet, and smart offices.

"We are tangibly and imperceptibly changing the business and efficiency of every client, and even more deeply changing the business paradigm and organizational structure of clients, which is what we hope to see," Zhang said.

Meanwhile, Zhang emphasized that because Zhipu has open-sourced multiple model versions, the global cumulative download volume of Zhipu ChatGLM-6B has exceeded 13 million.

Zhang also emphasized that the Zhipu AI team has secured the fifth place among the world's most popular open-source institutions, being the only Chinese institution and organization on the list.

Zhang admitted that as a company created by academics, Zhipu AI's strength in technological research and development is evident, but there are doubts about their commercialization. "Because what is needed in the market for technology in the laboratory is different, and the process is different."

Zhang believes that on the one hand, Zhipu AI's products must ensure originality and controllability, and on the other hand, "Zhipu's feature is to benchmark OpenAI's product line comprehensively. Of course, it's not to do what others do, but the concept is very similar to OpenAI's, aiming at General Artificial Intelligence (AGI)."

On the path of commercialization, Zhang remarked "we will see partners we cooperate with in various industries, focusing on different industries. We have transformed from a vertical chimney-like commercial development path into a development model that truly expects AI platform type, which is fundamentally different."

When talking about OpenAI's Sora, Zhang mentioned that Zhipu AI is also involved in related fields, but admitted that there is a gap. Therefore, Zhipu AI "was not unprepared." Since around 2021, Zhipu AI has been conducting related research on CogVideo. "Sora's capabilities have seen step-by-step growth, and we still have gaps and need to continue working hard."

"Internationalization is a very important strategy, everyone knows, perhaps some AI companies in the United States, its customers come from all over the world, not just limited to the United States. So from the perspective of us – a Chinese company, going global is definitely important," Zhang said.

转载请注明出处、作者和本文链接
声明:文章内容仅供参考、交流、学习、不构成投资建议。
想和千万钛媒体用户分享你的新奇观点和发现,点击这里投稿 。创业或融资寻求报道,点击这里

敬原创,有钛度,得赞赏

赞赏支持
发表评论
0 / 300

根据《网络安全法》实名制要求,请绑定手机号后发表评论

登录后输入评论内容

扫描下载App