AIGC Industry Weekly Report | Microsoft and Apple say goodbye to OpenAI's board observer seats; Damo Academy released a one-stop AI video creation platform "Seeking Light"; Sequoia United States Partner: AI will bring transformative opportunities to the service industry

Editor: Zhang Jinhe

Opening remarks: "AIGC Industry Weekly" sorts out the important developments of the AIGC industry in a week, product releases and the latest views of industry leaders.

Image source: Illustration

First, the dynamics of market enterprises

(1) Microsoft and Apple bid farewell to OpenAI's observer seats on the board of directors

Microsoft and Apple have waived their board observer seats for OpenAI, Microsoft said its limited role is no longer necessary, and OpenAI said it looks forward to continuing the cooperation and will no longer leave board observer seats for partners in the future.

The exit of the two companies may be due to antitrust pressures, and regulators have launched investigations into related cooperation, concerned that their monopoly position is hindering AI technology innovation and application.

OpenAI's path to a hybrid architecture is seen as a betrayal of its original intentions, and it could further transform into a for-profit enterprise and prepare for an IPO.

(2) OpenAI's CEO has founded another AI health company!

Altman and Huffington announced the launch of Thrive AI Health, Inc., with the goal of creating personalized AI health coaches.

The company hopes to improve people's living habits and prevent chronic diseases through AI, and the CEO is Love, the former head of health and wearables at Google.

Research collaborations have been established with a number of cutting-edge academic institutions and medical centers.

Second, product technology dynamics

(1) The web version of Coline is online

The web version of Kuaishou Keling is launched, the basic model is upgraded, new functions such as first and last frame control and lens control are added, and Wensheng video is open to 10 seconds, and at the same time, it can be open source.

Kuaishou has a large model matrix including Keling, Ketu, Kuaiyi, etc., and has built a full-process AIGC service for digital humans, and has also launched the agent "AI Xiaokuai" in the APP comment area.

Kuaishou adheres to self-research, embraces open source, deeply integrates large model technology with practical application scenarios, focuses on several major scenarios, and cooperates with universities to promote ecological development.

(2) SenseTime released Ririxin 5.5 and related products, and a number of evaluations exceeded GPT-4o

SenseTime released Ririxin 5.5, including the 5o version of streaming multimodal interaction, a number of evaluations of super GPT-4o, and also launched the device-side large model and related products.

SenseTime demonstrated the real-time audio and video interaction capabilities of RiRixin 5o on the spot, and also launched Vimi, an AIGC product that can generate controllable character videos.

SenseTime launched the 0 yuan Go plan to encourage developers and industry customers to join, with the price of end-side large models as low as 9.9 yuan per unit per year, and also launched a variety of industry large models and solutions.

(3) Damo Academy released the AI video creation platform "Seeking Light"

The Damo Academy released a one-stop AI video creation platform "Seeking Light" to solve the problems of poor controllability and cumbersome workflow in AI video creation.

The platform has layer-based video editing functions, which can improve creative efficiency, and the interaction is simple and the editing capabilities are rich.

"Seeking Light" will open for closed beta in the near future, aiming to become an exclusive video studio for creators and unleash the productivity of AI.

（四）Stable Diffusion 3允许商业化并将开源更大版本模型

Stability AI modified its community license agreement to allow businesses and individual developers with annual revenues of less than $1 million to use Stable Diffusion 3 Medium for free commercialization, which was previously only available for academic research.

Stability AI said that a larger version of the model will be released in the coming weeks and will continue to be open sourced, after SD3-M has greatly improved image quality, text semantic restoration, etc., and the training dataset has also been optimized.

Stability AI's ability to make timely changes to the protocol is well received, and its move to change the protocol will benefit developers and small businesses.

(5) Odyssey unveiled a new video model and raised $9 million in financing

Odyssey presents a video model with 4 built-in models that generate high-quality video elements and compose them into videos with Hollywood-grade effects.

The model supports two generation methods: splicing and text prompting, and the generated video can be exported as a 3D standardized format file for secondary editing.

Odyssey has a strong core development team and will work with Hollywood film and television production companies, and has already shown demo videos, which will be available soon.

3. Cutting-edge perspectives

(1) Report on the current state of artificial intelligence: applications, challenges and prospects

The report shows that sentiment towards AI has changed from cautious to slightly overestimated, with slow but steady progress in AI adoption, high adoption by small and large companies, and some people using AI in secret.

Among AI tools, OpenAI models are commonly used, the use of vector databases is increasing, most companies rent GPU resources from cloud providers, and HuggingFace is the most commonly used AI development tool.

The main obstacles to the development of AI applications include data security, etc., and people are highly satisfied with the AI technology stack, and most believe that the possibility of achieving general AI in the next decade is high.

(2) Pat Grady, partner of Sequoia United States, talks about the development and impact of AI

Pat Grady believes that AI technology is at a critical inflection point, and that a stable foundation model will help the development of the AI ecosystem, and that AI will bring transformative opportunities to the service industry and will not replace existing software companies.

He believes that the current model capacity is enough to build trillions of dollars of new business, the stability of the model is beneficial to the ecosystem, and the development of AI will shift from the training stage to the inference stage.

Pat Grady also mentioned the application of AI in venture capital and the empowerment of AI in the service industry, and believed that there is a funding bubble in the current AI field, but people have a clear understanding of the reality of its application.

(3) Zhang Peng, CEO of Zhipu AI, talked about the development of large models

Zhang Peng, CEO of Zhipu AI, believes that the implementation of large models requires cycles, and we should pay attention to the process of combining technology and application.

The goals and meanings of open source and closed-source models are different, and the commercialization focus of Zhipu is ToB, providing customers with a variety of solutions, and users have migrated from OpenAI, and the company has laid out international business.

The connotation of scaling law of large models is changing, and Zhipu believes that its development should move towards general artificial intelligence, and the next step is to have multi-modal capabilities to achieve "from virtual to real", and at the same time pay attention to security.

Source: Provided by Everyday Technology

National Business Daily