Navigating the Impact of OpenAI’s Sora on China’s AI Industry

By Giulia Interesse

As the world grapples with the implications of OpenAI’s Sora, a groundbreaking text-to-video AI, anticipation and caution grip the tech industry. In China, while domestic investors are optimistic, intensified competition and geopolitical tensions add complexity to the country’s AI ambitions. The Quarterly delves into Sora’s impact on China’s AI landscape, exploring opportunities, challenges, and the regulatory framework shaping the future of AI innovation in the country.

Robotic man cyborg face representing artificial intelligence concept 3D rendering

Sora’s debut marks a significant milestone in Artificial Intelligence Generated Content (AIGC), drawing global attention. While image generation has seen progress, video generation remained a challenge until now. Building on the success of OpenAI’s DALL-E, Sora represents a significant advancement, leveraging large language models (LLMs) to solidify OpenAI’s industry leadership.

In China, excitement, and concern surround Sora’s arrival. While it signals progress and anticipation for AI advancements, there’s also apprehension due to increased competition and potential constraints from US-China tensions. This article explores Sora’s emergence, analyzing its impact on China’s AI landscape, and the opportunities and challenges it presents for businesses, tech communities, and regulators.

What is Sora?

Sora, an AI tool, generates full-length videos, up to one minute long, based on user prompts, requiring only a simple input to produce a corresponding video. Despite its impactful capabilities, Sora’s rise may have gone unnoticed by those outside social media or specialized computing circles, arriving quietly without extensive advertising.

OpenAI has unveiled numerous example videos showcasing Sora’s remarkable skills, illustrating its talent in crafting lifelike animations. These examples exhibit intricate details like reflections in mirrors, precise fluid movements, and realistic simulations of falling snow particles, highlighting Sora’s impressive capabilities.

Sora’s Advantages

Sora marks a significant advancement in AI-driven content creation, particularly in video generation, surpassing text-based models like ChatGPT. It falls under the category of multi-modal large models, expanding the capabilities of GPT by effectively handling various video types. A key innovation lies in Sora’s treatment of video frames as sequences of patches, akin to word tokens in language models, enabling contextually relevant and visually coherent video generation based on text cues. Sora’s video training comprises three main steps: video compression network, spatio-temporal patch extraction, and video generation, resulting in remarkable improvements in both duration and resolution compared to previous models. It can generate 1080p resolution videos up to 60 seconds long, with the ability to edit and expand content based on text prompts. Moreover, Sora exhibits a deep understanding of text, accurately translating textual cues into detailed, emotionally resonant video content. It simulates physical laws, ensuring realistic three-dimensional motion and consistent object behavior across various perspectives. Unlike earlier models, Sora excels in maintaining coherence and consistency within videos, including elements like color style, facial expressions, and interactions between subjects and backgrounds, enhancing overall quality and realism.

Sora’s Applications and Uses

Sora’s versatility opens diverse applications across various industries, including:

  • Creative fields: Filmmakers, visual artists, and designers can utilize Sora to innovate their expressions, generating visuals directly from scripts, simplifying conceptualization and pre-production.
  • Education and training: Sora aids in crafting intricate educational materials, including historical recreations and scientific simulations, enhancing learning with engaging, immersive content.
  • Advertising and promotion: Sora enables businesses to create visually captivating video content for marketing campaigns based on textual descriptions, streamlining content creation and fostering innovative advertising.
  • Gaming and virtual reality: Developers can integrate Sora to enrich gaming and virtual reality environments, contributing dynamic backgrounds, character interactions, and cutscenes, elevating narrative elements and overall user experience.

Sora Prompt: A stylish woman walks down a Tokyo street filled with warm glowing neon and animated city signage. She wears a black leather jacket, a long red dress, and black boots, and carries a black purse. She wears sunglasses and red lipstick. She walks confidently and casually. The street is damp and reflective, creating a mirror effect of the colorful lights. Many pedestrians walk about.

Photo courtesy of Open AI

How is China Viewing Sora?

Discussions surrounding Sora in China have triggered a range of reactions, from admiration to anxiety about potential lagging behind in AI advancements. Within China’s dynamic business and tech circles, there’s a mix of anticipation and apprehension surrounding OpenAI’s latest breakthrough. The announcement of Sora’s arrival has been viewed positively by domestic investors, with the Sora Index, reflecting 49 technology, entertainment, and media companies, surging by 11.4 percent post-Chinese New Year holiday. Generative AI’s development is seen as promising, driving growth in computing power globally and in China, as per Ping An Securities.

However, amidst this excitement lies the backdrop of increasingly fierce competition within the AI sector and the persistent challenges stemming from the ongoing US-China trade tensions. Observers have raised concerns about potential “supply chain risks” should the US continue to restrict chip exports to China. While these sanctions might catalyze the development of the domestic AI chip industry in China, the pursuit of alternative solutions could demand considerable time and effort to meet the high expectations set by the evolving landscape.

Does China have its Own Sora?

The Chinese government, through the National Cyberspace Administration (CAC), mandates registration for all publicly accessible large language models (LLMs), while foreign players like OpenAI and Google have no direct presence in mainland China, although Microsoft’s Copilot, utilizing OpenAI’s GPT models, is accessible in Hong Kong. This absence has fueled local tech giants like Baidu, Tencent Holdings, and Alibaba Group Holding to aggressively compete in a market crowded with over 200 LLMs, with companies like Cloudwalk Technology focusing on enhancing text-to-image and text-to-video functionalities, unveiling a “digital human” generation platform. Similarly, Sumavision has invested in video content production and is dedicated to further exploring AI-generated content technologies.

These strategic moves by Chinese AI firms reflect a broader trend in leveraging advancements in video generation technologies and expanding digital content offerings. However, few can match Sora’s capabilities, with ByteDance, the parent company of TikTok, acknowledging the early stage development of its internal video motion control tool, Boximator. To bridge the gap, Chinese innovators must prioritize technology research and development, talent cultivation, and market insights while collaborating with established technology teams, sharing technical resources, and tailoring products to domestic market needs.

Rather than directly competing with Sora, some industry insiders focus on gaining access to OpenAI’s model. Beijing-based Sinodata aims to be among the first to apply for a Sora API subscription once available on Azure, suggesting a preference for embracing foreign expertise and collaboration to advance their LLMs, fostering opportunities for mutual growth and innovation in the AI landscape.

Regulation of Sora-like AI Applications in China

China has emerged as a leader in AI regulation, with initiatives like Made in China 2025, the Action Outline for Promoting Big Data Development, and the Next Generation AI Development Plan fostering AI sector growth while establishing ethical guidelines for AI businesses and algorithms. The government also prioritizes the expansion of its data economy, emphasizing infrastructure development to support basic data systems. In efforts to regulate the AI sector, China introduced the Interim Administrative Measures for Generative AI Services in July 2023, covering various aspects of generative AI service creation and delivery. These measures highlight China’s commitment to advancing AI development while ensuring ethical practices. By creating a secure operational environment, China aims to encourage innovation while monitoring the ethical application of AI across sectors.

Opportunities in China’s AI sector

China’s AI sector offers promising investment prospects and is a key focus area for the government, despite regulatory hurdles. Efforts are underway to create a favorable environment for foreign investment and talent in the industry, with projected AI investment in China expected to reach US$26.69 billion by 2026, comprising about 8.9 percent of global investment. China’s vast market, growing consumer base, and advanced infrastructure further augment opportunities for companies to develop and expand their AI offerings.

To succeed in China’s evolving AI market, businesses need to capitalize on supportive policies and align with governmental objectives, leveraging available incentives to unlock various opportunities. Maintaining adaptability is crucial for navigating the dynamic business landscape and political dynamics inherent in the competitive AI market.


This article is from the AmCham China Quarterly Magazine (Issue 1, 2024). To access the entire publication for free, sign up on our member portal here.