Soon after the beginning of 2024, the AI world has ushered in a huge change and a new round of wrestling.
In the early morning of February 16, OpenAI released the Wensheng ** model SORA on its official website. Users only need to enter text instructions, and SORA can quickly generate 60s ** content, while understanding and simulating the real world, and output ** that accurately reflects the user's prompts.
In the eyes of some industry observers, SORA is not only far ahead of its competitors, but can even be considered a "simulator of the world". In this context, the artificial intelligence industry will usher in a new round of SORA in this revolutionary technology, and the feedback to the capital market will be reflected in the SORA concept stocks detonating AI at home and abroad, and on February 19, SORA concept stocks set off a rising tide in the A** field.
What is more concerning than this is that the investment opportunities in AGI-related industries have not yet seen the ceiling, and there is no doubt that there are explosive growth opportunities from them.
SORA is leading the technological change, and 2024 may be the year of AI.
From Apple's release of the Vision Pro headset at the beginning of the year, to the successive release of AIPC by major PC manufacturers, and then to the release of SORA by OpenAI this time, the world's innovation in artificial intelligence is accelerating, and the iteration is getting faster and faster.
OpenAI positions SORA as a generative model for world simulators, laying the foundation for future development to understand and simulate real-world models. According to the official introduction, Sora currently has the ability to generate subjects and complex scenes that include multiple characters, specific types of movements, accurate details, and so on. At the same time, it is also possible to create multiple shots to simulate the complex camera movement effect, making the generated ** more vivid.
In addition, because OpenAI combines diffusion models with LLMs, SORA has the ability to understand and simulate the real world, so the ** generated by SORA can already show the real physical relationships between space and objects in space. At present, in addition to the Wensheng function, SORA also has the functions of picture generation, extension, splicing, editing, image generation and so on.
This will undoubtedly be a huge change in the field: from the perspective of the product and technology itself, the speed of human generation has been increasing in the past 10 years, but from 1 second to 4 seconds, it took 10 years to make a breakthrough. Even tools such as pika and runway are stuck on the limit of more than ten seconds.
However, OpenAI's SORA directly raises this limit to 60 seconds, and compared with previous AI Wensheng tools such as PIKA and Runway, SORA not only has the ability to complete more complex tasks such as forward expansion and splicing, but also brings more vivid multi-perspective through multiple lenses and other methods, and at the same time has significant advantages in terms of duration, fluency and logic in generating.
Source: Huatai**.
ICBC Credit Suisse research report also pointed out that SORA has adopted the first-class generation large model technology route of "diffusion + transformer", which can be described as a milestone technological progress. Compared with the U-NET, which is also a diffusion model used in the past, the parameters of the Transformer architecture are highly scalable, that is, the number of parameters is increased, the performance is accelerated, and the training data of any resolution, aspect ratio, and duration is supported, and the quality of the training data will not be degraded due to compression. In addition, SORA is trained on an autoencoder that can compress ** in time and space, which is why it can greatly increase the generation time.
The great success of the SORA model marks the iterative upgrading of AI applications, which is expected to greatly shorten the expected time of AGI general artificial intelligence to a certain extent, and bring a huge wave of change from technology to the whole industry.
Wensheng** iteration accelerated, and the growth potential of Global Mexico was highlighted.
From a commercial point of view, SORA is expected to profoundly change the major subdivisions, among which the pan-entertainment track is more than expected. Because the realism of the spatial details generated by the SORA model exceeds many people's expectations, it will have great application potential in the fields of image creation and creativity in the future.
It is foreseeable that the application of multimodal models will usher in explosive growth in 2024, and enterprises that take the lead in laying out related fields are expected to run out of "new" giants.
From the perspective of the versatility and scenario application of SORA, the content generated by AI automatic creation will affect many industries, or will bring a new round of content creation industry revolution. However, according to the underlying logic of ** generation, SORA unifies the data representation of images and **, and expands the scale of the model and improves the performance of the model through a large number of datasets, and players in the industry with a large number of high-quality images or ** resources are expected to be the first to benefit.
If you look for targets under this logic, Global Murphy (GMM., a 3D content company listed on the U.S. stock marketUS), which is undoubtedly expected to accelerate in this wave of technology-led change.
Zhitong Financial APP has learned that Global Murphy is a company with digital technology as the core and provides customized services for the metaverse industry by developing innovative solutions. The company has been focusing on 3D visualization, adhering to the "AI+3D" technology development direction, combining virtual digital assets with real-world interactive experience, breaking the boundaries of traditional media, and providing customers with rich visual effects and immersive experiences.
The Company creates a wide range of 3D high-definition virtual versions of physical world objects (such as people, animals, and scenes) that can be used in different applications, primarily through its proprietary "MofyLab" technology platform, which consists of cutting-edge three-dimensional (3D) reconstruction technology and artificial intelligence (AI) interaction technology. As of now, Global Murphy is one of the leading digital asset banks in China, consisting of more than 30,000 high-precision 3D digital assets, which are expected to become important underlying assets in the field of digital content and generation.
It is worth noting that in recent years, Global Murphy has been actively deploying in cross-modal fields such as Wensheng images and **. In the company's 3D digital asset business, involving a large number of 3D modeling, 3D rendering and IP content asset generation, the company has been committed to presenting cutting-edge technology and visual experience to customers, so at the AI application level, the company has been actively promoting the landing of 3D model + AI and scene application, as well as the visual interactive experience of AIGC.
With the decline in the cost of 3D content, in the process of promoting digitalization and technological progress in China, there will be more scenarios to combine emerging technologies such as SORA to achieve the rapid implementation of metaverse applications, thereby generating demand for some content services and bringing incremental performance to Global Murphy. If we superimpose the era of spatial computing, scenarios such as VR and MR also need AI 3D models to expand and enrich the content supply of the virtual reality industry.
Looking forward to the future, under the wave of breakthroughs in AI Wensheng technology led by SORA, multimodal fields such as 3D are expected to continue to accelerate, continue to improve the generation efficiency and accuracy of AI 3D models, and promote the arrival of the inflection point of AI 3D engine technology. Global Murphy is expected to maintain its leading edge in digital assets, connect the virtual and real world through modeling, ** and other technologies, provide data-driven and model-based solutions for the industry, and enjoy greater growth dividends as a beneficiary of the underlying technological change and a content material provider.