As we look ahead to 2025, it is clear that AI technology will play an even more pivotal role in the industrial landscape.

Large AI models will continue to integrate seamlessly into daily business processes, driving the evolution and upgrading of enterprises and entire industries alike.

Author | Dou Dou

Editor | Pi Ye

Produced by | Industrialist

Recently, OpenAI's '12-day serial drama launch event' came to a successful conclusion.

With the rapid advancement of AI technology, the demand for AI in the industrial sector is growing, particularly in terms of improving efficiency, reducing costs, and enhancing competitiveness. In a sense, OpenAI's 12-day live launch event was not just a technology showcase but also a profound glimpse into the future of industrial transformation.

Features like reinforcement fine-tuning, Sora Turbo video editing, the Python runtime environment on Canvas, and AI desktop assistants directly address these needs, providing robust support for industrial implementation.

For instance, reinforcement fine-tuning technology enables significant performance improvements even with limited data, leading to lower inference costs and faster knowledge base construction for enterprises with constrained data resources. This not only lowers the barrier to AI adoption for enterprises but also offers the potential for rapid response to market changes.

Another example is the video editing capabilities of Sora Turbo, which provides new creative and editing tools for the media and entertainment industries, enhancing the flexibility and efficiency of content production.

Additionally, the Python runtime environment on Canvas lowers the barrier to programming, enabling non-technical personnel to quickly get started, thereby accelerating technology application and innovation. The highly acclaimed AI desktop assistant enhances workflow smoothness and intelligence by directly collaborating with local applications.

The development and application of these technologies not only highlight the rapid advancement of AI technology towards AGI but also signify that industrial AI may become a focal point for AI in the coming years.

1. Vertical Models Remain the 'Main Channel'

In the realm of AI, the debate over general-purpose AGI versus vertical domain models has been ongoing. Over the past 12 launch events, OpenAI has provided some insights.

In recent live streams, OpenAI demonstrated its reinforcement fine-tuning technology. Unlike traditional fine-tuning, this technology enables high-quality AI models to be quickly obtained even in vertical domains with limited data. Ultraman hailed this technology as one of the biggest surprises of 2024.

As a manufacturing powerhouse, China's enterprises possess vast amounts of industry data, providing abundant 'nourishment' for the development of AI.

While China has established a solid research foundation in the AI field, especially in computer vision and speech recognition, and has made significant progress in applications, there is still a gap with international advanced levels in basic theory and original algorithms. Especially in terms of original algorithms and model architectures, there is room for improvement.

For instance, OpenAI recently released a faster and smarter full-featured o1 model and launched the most expensive o1pro, costing up to $200 per month. Through model optimization, it also introduced the full-featured o1, reinforcement fine-tuning, and the o3 family.

In particular, the o3 series claims to be close to general AI. OpenAI stated that o3 scored 87.5% on the ARC-AG test, surpassing GPT-3 and GPT-40. It scored 2727 on the programming competition Codeforces and achieved a 96.7% accuracy rate on the math benchmark AIME 2024.

OpenAI's model optimization underscores the potential of its AI technology in processing speed and intelligence level. For the development of AI technology in China, this implies a need for continuous investment in algorithm innovation and model training.

However, this shortcoming cannot be remedied quickly.

The main reason is the insufficient investment in AI basic research in China, leading to fewer original achievements and reliance on foreign research progress. Additionally, data resources are scattered across different enterprises and institutions, lacking an effective sharing mechanism, which also limits the effectiveness of model training.

Nevertheless, with the deepening of digital transformation, various industries have an increasing demand for intelligent solutions. Vertical domain models can quickly respond to market demands, empower various industries by incorporating specific industry characteristics, and drive industrial upgrades. This will not only propel the application and development of AI technology in China but also potentially achieve overtaking in certain fields.

2. AI is Moving Towards Integrated Listening, Writing, and Viewing

In March 2024, OpenAI launched its video model Sora, marking a new era in video generation technology. This move not only sparked a positive response and pursuit from domestic manufacturers but also heralded a new chapter in the development of multimodal technology.

During OpenAI's 12-day live stream, it was upgraded again, releasing the official version of Sora, which supports generating videos up to 1080P resolution and up to 20 seconds long, with support for multiple video aspect ratios.

More importantly, Sora Turbo was also launched. The highlight of Sora Turbo is its innovative storyboard function, allowing users to edit videos from any time point, breaking the limitation of traditional video models that can only generate single videos and enabling the creation of complex video sequences.

Currently, OpenAI has stated that Sora is only available to ChatGPT Plus and Pro users, with the former receiving a monthly quota of 50 video generations and the latter up to 5,000.

This feature significantly enhances the accuracy and personalization of video creation, enabling creators to express their creativity more freely.

Simultaneously, OpenAI integrated video chat and screen sharing functions into its advanced speech mode, realizing real-time interaction between vision and hearing, further enriching the user communication experience.

These two upgrades jointly promote AI's capabilities in multimodal creation, making the conversion from text to video more efficient and intuitive. By integrating speech, vision, and text, intelligent assistants like ChatGPT can not only better understand and respond to human needs but also provide more comprehensive support when handling real-time tasks.

This advancement in multimodal technology not only improves the quality of human-computer interaction but also provides unlimited possibilities for cross-domain application development.

The development of multimodal technology is not just a technical breakthrough but also reflects a deep understanding of human cognition and interaction methods.

A new revelation is that future AI development should prioritize humanized design to meet people's increasingly complex and diverse practical needs. As technology continues to evolve, we can foresee that future interactive interfaces will integrate hearing, vision, and text, forming a more natural, intuitive, and efficient communication environment.

With innovative technologies like Sora Turbo emerging, AI is rapidly moving towards integration of listening, writing, and viewing, bringing unprecedented transformation opportunities to various industries.

3. Large Model Enterprises Have the Responsibility to 'Build Bridges' for AI Applications

During OpenAI's 12-day live launch event, it was evident that a series of new features and tools, such as the free opening of the Canvas canvas function, the launch of the Project function, and the debut of the AI desktop assistant, demonstrate the company's efforts to expand the boundaries of AI technology.

This not only signifies the progress of AI technology itself but also reflects its potential to profoundly impact various industries.

Specifically, the free opening of the Canvas canvas function creates a new platform supporting Python programming and transforms into a multifunctional AI tutor through the integration of GPT intelligent assistance. This not only lowers the technical threshold for programming and creation, enabling more people to participate in technological innovation but also brings revolutionary changes to the fields of education and technology development.

The launch of the Project function further enhances GPT's capabilities in project management based on user feedback.

It supports users in consolidating information such as materials, files, and chat records into a single Project, supporting scenarios such as project management, writing, file and data management, and personalization. In short, it helps users plan, organize, and complete projects more efficiently and increases work efficiency and project success rates through intelligent analysis and task management.

OpenAI stated that it plans to provide this to enterprise and education users early next year.

The debut of the AI desktop assistant not only directly interacts with local applications but also significantly enhances workflow smoothness, making daily work more intelligent and convenient.

It is understood that based on the AI desktop assistant, users can collaborate with applications such as Warp and Xcode through simple copy-and-paste operations, executing tasks without detailed communication. Additionally, ChatGPT supports collaboration with applications such as Notion and Apple Notes in voice mode. It is currently available on the latest version of Mac and in the ChatGPT application.

In summary, the integration of the above functions provides an efficient and open innovative environment, stimulating broader application possibilities and promoting the in-depth application of AI technology in different fields. For the domestic and even global AI industry, this also provides valuable reference significance on how to transform advanced technology into actual productivity.

Looking ahead, with the continuous emergence of similar innovative platforms and the development of project management tools towards intelligence and automation, it is expected to trigger a revolution in working methods, significantly improving work efficiency and professionalism.

This evolution is not limited to the technical level but will also profoundly change people's work patterns and collaboration methods, propelling society towards a higher level of the information age.

4. The Premise of AI Technology Democratization: Lower Thresholds

Achieving technology democratization is crucial for promoting technology implementation, and to achieve this, the first step is to lower the thresholds for using these technologies.

Against this background, progress in the AI field is particularly notable.

During the OpenAI launch event, the debut of the o1 model API became a new focus in the developer community. The newly added WebRTC support enables real-time voice interaction with just 12 lines of code, reducing costs by 60% and greatly simplifying the development process of AI applications.

Simultaneously, preference fine-tuning tools were added, allowing developers to customize AI models according to specific user needs, providing a more personalized user experience.

These two features significantly simplify the AI application development process, making the creation of complex functions more direct and convenient.

With the opening of APIs, the technical threshold is further lowered, stimulating the boundless creativity of developers. They can now more easily build efficient and innovative AI solutions, injecting strong momentum into the rapid popularization and development of AI technology.

The opening of APIs is not only an important indicator of AI technology democratization but also opens the door to more developers and innovators, enabling them to access the most advanced AI models and tools, thereby accelerating the pace of innovation in the entire industry.

The significance of this open strategy goes beyond that. It promotes technology sharing, encourages a wider range of creative ideas, and spawns diverse solutions. This not only accelerates industry development but also allows more small and medium-sized enterprises and individual developers to participate in the development and innovation of AI technology. Ultimately, this trend will bring a rich variety of AI applications and services to society, truly realizing the popularization and democratization of AI technology.

5. 2025: Exploring the Infinite Possibilities of AI Seamless Integration

If there is one direction with the highest product concentration in OpenAI's 12-day launch event, it is the various 'means' for consumers to access AI. Examples include free search services, deep integration with the Apple ecosystem, and new ways to communicate with GPT through multiple channels.

Among them, OpenAI launched a global free and precise search function based on a fine-tuned version of the GPT-40 model. By combining content directly provided by third-party search providers and ChatGPT partners, users can quickly and accurately obtain the required information. The built-in map and support for advanced voice modes provide users with a brand-new search experience.

The launch of this feature not only eliminates advertising interference but also enables users to obtain information more easily and quickly through advanced voice modes with its intelligent speech recognition capabilities.

Simultaneously, ChatGPT is fully integrated into the Apple system, supporting Siri, camera control, and shortcut operations, which not only significantly enhances the advantages of the Apple ecosystem but also brings a richer interactive experience to users.

Currently, iPhone, iPad, and Mac users can use ChatGPT functions through Siri. Users of the entire Apple ecosystem can now more conveniently interact with AI and enjoy the convenience it brings.

Furthermore, to enable more people to benefit from AI advancements, any phone or mobile device with dialing capabilities can directly communicate with GPT, and WhatsApp users can also interact with GPT through messages. It is understood that the number is 1-800-CHAT-GPT or 1-800-242-8427, currently supporting smartphones, feature phones, landline phones, and other devices.

This diversified communication approach significantly lowers the entry barrier, making AI technology more accessible and user-friendly, thereby truly integrating it into thousands of households.

Whether it's the introduction of free search services, deep integration with the Apple ecosystem, or various GPT communication methods, these have all notably enhanced the user experience, demonstrating that AI technology is progressively becoming an indispensable aspect of our daily lives. These seamlessly integrated services not only foster a more natural and intuitive human-computer interaction but also elevate the quality of life and work efficiency, ushering in a smarter future.

Postscript:

As OpenAI's 12-episode serial drama launch event draws to a close, we have observed AI technology seamlessly integrating into our lives and work at an unprecedented pace and depth. From the sophisticated development of vertical models to innovative advancements in multimodal interaction, and the democratization and seamless integration of AI technology, each step signifies a monumental leap in the realm of artificial intelligence.

Looking ahead to 2025, it is foreseeable that AI technology will assume an even more pivotal role within the industry.

AI technology will further embed itself into the daily workflows of enterprises, establishing itself as a benchmark for enhancing productivity and efficiency. As technology matures and costs decline, enterprises will increasingly turn to AI to refine decision-making, elevate service quality, and augment customer satisfaction. The seamless integration of AI technology will empower enterprises to adapt more agilely to market fluctuations and promptly address customer demands.

Concurrently, AI will propel the industry towards a more intelligent and automated trajectory, presenting unprecedented growth opportunities for enterprises.