Sora Advances: Will Artificial Intelligence Transform the Future of the Sports Industry?
-
On February 15th, U.S. time, OpenAI unveiled Sora, a large-scale text-to-video model. By inputting brief descriptive text, Sora can generate a one-minute video featuring multi-angle shots and elements that interact in accordance with real-world physics, marking a significant technological leap compared to other AI video demonstrations from 2023.
Since OpenAI introduced ChatGPT at the end of 2022, breakthroughs in artificial intelligence have reached a level where they can be preliminarily applied in everyday public domains. Now, with the emergence of Sora, the evolution of AI's learning capabilities is further demonstrated, suggesting that the advent of the AI era across various industries may not be far off.
As a multi-billion-dollar entertainment industry, the sports world is also highly attentive to the potential applications of artificial intelligence. Coincidentally, a year ago in February 2023, during the sports business-themed dialogue program The Community co-produced by China Voice's Decisive Moment and Sports Big Business, the possibility of AI revolutionizing sports was discussed. Guest speaker Liu Jianhong, Chairman and CEO of Kelly Technology, introduced existing AI applications such as automatically generated marathon runner videos and intelligent live broadcasts of table tennis matches. Now, with the debut of Sora, what new AI-driven possibilities does it bring to the sports industry? From the perspective of Sora's text-to-video capabilities, its applications in the sports industry are relatively limited. This is due to two major conflicts between the characteristics of sports and Sora's intended use cases.
First, Sora produces fictional video content, whereas the sports industry offers products featuring real athletes engaged in actual competitive events. From a media perspective, viewers of sports videos seek highlights and broadcasts of games that have occurred in objective reality. Sora, which creates content in a fictional format, cannot provide similar products.
Second, the content generated by Sora is primarily controlled by the "prompts" provided by the creator, while sports are characterized by unpredictability and randomness. From a narrative standpoint, Sora's output is constrained by the creator's guidance, with prompts defining the content outline. In sports, however, the athletes—acting as "creators"—rely on their physical abilities as their core creative tools. There is no predefined "outline" that dictates an athlete's actual performance. This unpredictability is one of the classic charms of sports, which Sora also fails to replicate. Both Sora and the GPT model behind ChatGPT are based on the Transformer architecture. For general users, the experience of using both is similar to a 'command-response' process. However, the similarities between Sora and GPT do not imply that their applications in the sports field are equally broad. GPT is highly effective in organizing, mining, and analyzing data, which is a crucial aspect of the sports industry. Whether it's predicting match win probabilities, uncovering interesting match statistics (e.g., identifying the youngest player to score in consecutive matches), or analyzing athlete/team performance, GPT has a wide range of applications.
GPT assists in analyzing data from the German second division and minor tennis tournaments. Source: StatsPerform
GPT primarily generates text and images, which can help sports associations and event organizers save time by automatically producing standardized post-match reports. In the media industry, GPT can aid journalists in writing articles and editors in crafting headlines, benefiting sports media as well. Sora may only provide services in a few specific areas related to sports videos. The author envisions two main scenarios. One is the optimization of sports program visuals. Beyond text, Sora can also use images and videos as prompts, giving it powerful video expansion capabilities. Creating program intros and transition effects is naturally within its scope, but Sora can also tackle more advanced visual presentations.
For example, sports broadcasts often include atmospheric shots, such as stadium exteriors or close-ups of natural phenomena like sunsets or snowfall. Sora can produce grand, stylized visuals to complement these atmospheric shots, enhancing the event's epic scale. Imagine starting the game footage with a dazzling Milky Way effect, then rapidly zooming in from the galaxy to the solar system, then to Earth, to the host city, and finally to the stadium—couldn't such visuals further amplify the audience's excitement? Sora could also render athlete entrances as "superhero transformations," seamlessly blending self-generated animations of athletes surrounded by cool elements like wind, fire, or electricity with actual footage of them preparing for competition. For instance, sports broadcasting began experimenting with 360° panoramic views a decade ago. This technology allows viewers to adjust perspectives seamlessly from any angle during any moment of the game. Traditionally, creating such footage required multiple cameras arranged in a circular array, followed by complex post-production algorithms to synthesize footage from each camera. The high cost of equipment made panoramic production prohibitively expensive, limiting its use to major tournaments. However, with Sora's involvement, its self-learning capabilities could potentially reconstruct accurate object trajectories from limited camera angles by analyzing morphological differences of the same object across different views, ultimately generating highly realistic continuous footage. This would enable budget-friendly panoramic effects that enhance broadcast appeal.
Sora's second potential sports application lies in fan engagement. Event organizers and IP owners could provide video generation tools allowing fans to upload photos and create videos of themselves competing alongside star athletes. Leveraging Sora's generative capabilities, each video would be truly unique. Theoretically, the current version of Sora can produce similar videos without requiring many third-party plugins. However, there is considerable debate in the industry regarding the copyright of information sources used during AI self-learning. The portrait rights of real athletes and the intellectual property of actual sporting events obviously cannot be used freely. This means, from a regulatory standpoint, virtual video production services—such as fans appearing alongside stars—can only be provided by event organizers, talent agencies, and the athletes themselves.
Although the sports industry may not have an immediate, obvious need for Sora's functionalities, the significance of Sora extends beyond just video production. Its breakthrough lies in reflecting the progress of AI self-learning and demonstrating AI's ability to absorb and understand a vast array of specialized knowledge.
Precisely because of the technological breakthroughs embodied by Sora, as society contemplates the future applications of AI, there is also concern about industries being replaced by machines. Some jobs may become permanently obsolete, and entire industries could face extinction. The sports industry inevitably faces similar contemplations. Will artificial intelligence fundamentally alter the face of sports? Could machines eventually occupy positions within the sports sector? The answer could be "yes," but ultimately, it's likely "no."
The notion of "yes" primarily stems from the certainty that AI's utilization scope and adoption rate will undoubtedly increase in the foreseeable near future. Taking the project introduced by Liu Jianhong in the "Community" program as an example, the automatic generation of marathon runners' competition videos represents a new service made possible by advancements in AI. Without AI involvement, no event organizer could provide personalized race videos for tens of thousands of participants, demonstrating AI's deepening integration into sports. Meanwhile, the automated broadcasting of table tennis events shifts tasks that previously required human supervision to AI systems, exemplifying improved AI utilization rates and serving as a classic case of machines replacing human roles. Looking further into the distant future, if artificial intelligence becomes a universally adopted tool applied across various aspects of daily life and work, entirely new forms of sports based on AI will emerge. In a 2023 article discussing the development of esports in Hong Kong, China, the author proposed that both traditional sports and video games are forms of "game" that will diversify into rich subtypes as productivity evolves. When AI combines with cutting-edge technologies like VR, AR, and holography to form a new productivity toolkit, future generations will invent sports forms beyond our current imagination - just as people from two centuries ago couldn't comprehend motorsports, esports, or drone racing. The disruptive potential of AI will become even more apparent then. Imagine future sports venues being dynamically created by AI in real-time?
The varying degrees of AI development in both near and distant futures collectively create possibilities for machines replacing humans and AI disrupting traditional sports. However, this doesn't mean humans will be absent from future sports worlds, nor does it imply the disappearance of jobs in the sports industry. The key point is that machines cannot replace humans as the main body in sports.
This statement can be explained from the perspective of biological evolution. There is a classic debate in the field of artificial intelligence: whether simulating brain activities or simulating physical movements is more difficult. The impressive achievements of models like GPT and Sora demonstrate rapid progress in simulating brain activities. In contrast, the development of bionic robots lags far behind. Robots' walking, running, and jumping movements are clumsy and rigid, let alone the instinctive ability of natural animals to adapt their limbs to terrain and state changes, smoothly completing daily movements.
From an evolutionary standpoint, scholars have proposed a hypothesis: Earth's animals took hundreds of millions of years to evolve from single-celled organisms like paramecia to humans. The emergence of primates with relatively developed brains and the ability to use tools to some extent occurred only tens of millions of years ago. This suggests that the iterative scale of human limb functions far exceeds that of brain evolution, representing a highly difficult and prolonged process. When humans seek to simulate themselves with machines, the achievements in simulating the brain are correspondingly easier to achieve than simulating physical movements.
Based on this conjecture and the slow development of bionic robotics, we can further deduce that human comprehensive limb utilization represents an extremely advanced form of movement. The pursuit of physical coordination, flexibility, stability, and strength in sports constitutes an even higher-level art within these advanced movement forms. If machines struggle to simulate basic human limb movements, they are even less capable of replicating the ultimate beauty of human athletic performance. While various industries face challenges from artificial intelligence, sports possess unique characteristics that prevent complete replacement by machines.
Therefore, the sports industry's fundamental attitude toward AI should align with researchers' initial approach—viewing it as a new tool to enhance efficiency and improve outcomes. Just as researchers encounter technical, ethical, and legal obstacles in expanding AI applications, the sports industry will inevitably undergo changes in presentation methods and creative approaches due to AI's evolution. However, just as researchers persist in technological breakthroughs, the sports industry will discover new vitality through embracing artificial intelligence.