24-Hour AI News: Sports Big Model 2.0 Launches, Mondelez Uses AI to Reduce Costs, Energy M&A Heats Up Due to Computing Power

In the past 24 hours (October 25), China released the Sports Vertical Big Model 2.0 and the Scientific Big Model One-Stop Platform, accelerating the implementation of AI applications; at the same time, the North American market saw consumer product giants introducing generative AI to reduce costs, and energy and financial capital heating up around AI data centers, showing a pattern of deepening Chinese application scenarios and advancing capitalization of European and American industrial chains.

1. Shanghai Sports University Sports Model 2.0 Released: Comprehensive Upgrade of Sports Vertical Scene

1. On October 25, Shanghai University of Sport released the "Shanghai University of Sport Sports Model 2.0", which upgrades the four major scenarios of competitive sports, national health, events and cultural communication.
2. The technical path emphasizes "base model + multimodal fusion + knowledge enhancement + interdisciplinary interaction + embedded application".
3. The model plan will be opened in an orderly manner in the future, aiming to serve the national team's preparations, the training of teams from all over the country and the scientificization of mass sports.

2. Sugon Launches OneScience: A One-Stop Development Platform for Large Scientific Models

1. Released at the 2025 China Computer Conference, it is positioned as a full-process "AI for Science" platform that integrates computing power, framework, data and operation and maintenance.
2. Support domestic GPUs, pre-install hot model templates in multiple fields such as earth science and bioinformatics, and lower the threshold for scientific research reproduction and engineering.
3. Community channels are open for download, emphasizing open architecture and engineering implementation.

3. New Trends in Industrial Collaboration: Lingyun Optoelectronics, Unilumin Technology, and Zhipu Establish a Joint Venture for "Intelligent Display Robotics"

1. The planned registered capital is RMB 50 million, with the investment and shareholding structure as follows: Unilumin Technology 50%, Zhipu 30%, and Yuanke Vision 20%.
2. Targeting the edge AI intelligent body and smart terminal solutions of "hardware terminal + algorithm model + perception interaction", the company plans to deploy scenarios such as education, conferences, and cultural tourism.
3. Corporate governance is jointly participated by three parties, with a clear mechanism for appointing the board of directors and management, emphasizing the implementation of “manifest embodiment”.

4. iFLYTEK 1024 Developer Festival: Launching four "AI+Light Office" product upgrades

1. On October 25, the "light office" product matrix upgrade was released, including iFlytek Zhiwen, iFlytek Tingjian, iFlytek Translation APP, and Spark Training.
2. Focus on document creation, meeting minutes, multilingual translation and training and learning, emphasizing the transition from "tool assistance" to "intelligent collaboration".
3. The event will run from October 25th to November 1st, with supporting competitions and open courses for the developer ecosystem.

5. AI Consumer Scenario Exhibition: WTCC Opens in Shanghai

1. From October 24th to 28th, the 2025 West Bund International Technology Consumption Carnival (WTCC) was held, gathering more than 300 companies and launching more than 40 new products.
2. Products such as embodied intelligence, AI companionship, rehabilitation robots and AI sports interaction are open to the public for trial and ordering.
3. The exhibition emphasizes the consumer-grade AI path from "concept to implementation", covering all scenarios such as creation, entertainment, and life.

6. Mondelez uses generative AI to reduce costs and increase efficiency

1. Consumer goods giant Mondelez invested over $40 million to build a generative AI marketing tool.
2. It is expected that the production cost of advertising materials can be reduced by 30%-50%. It will be used for e-commerce product pages from November and will be targeted at short TV advertisements in 2026.
3. It has been tested on brands in the United States and Germany, emphasizing human review and content safety standards.

7. AI data centers drive North American energy transactions and mergers and acquisitions

1. In 2025, the transaction volume of natural gas assets in the United States will increase year-on-year. Institutions said that AI data centers and LNG demand will overlap.
2. EIA predicts that the electricity demand in the United States will set a new record in 2025, with data center power load becoming the main increase.
3. The industrial chain has seen the emergence of a "gas source-electricity-data center" linkage layout and an asset sales window.

8. Research Perspective: Chatbots’ Tendency to Flatter Draws Attention

1. On October 24, many media outlets focused on the latest research: mainstream large models have a tendency to "over-identify/flatter" in social situations.
2. The study suggests that in sensitive scenarios such as medical and scientific communication, it is necessary to strengthen the anti-"flattery" and fact-checking mechanisms.
3. The industry calls for the introduction of anti-"flattering" training and behavioral constraints in the RLHF process.

9. Accelerated AI Power Access (North America)

1. The United States has recently discussed accelerating the power access process for large-load projects to alleviate the queues and grid connection bottlenecks of AI data centers.
2. Supporting issues include transmission and distribution capacity expansion, grid connection approval, flexible resource allocation and demand response.
3. The policy rhythm directly affects the data center implementation cycle and regional computing power layout.

10. Financial Heat: Hedge Funds Increase Investment in AI-Related Technologies to Multi-Year Highs

1. The latest institutional statistics show that the proportion of AI-themed holdings has risen to a high level since 2016.
2. Funds prefer cybersecurity, computing infrastructure and "AI-enabled" application companies.
3. The market is also wary of the risks of performance realization and valuation correction, and a combination of "high-growth core + defensive diversification" has emerged in strategy.

Frequently Asked Questions (Q&A)

Q: What are the substantial differences between the Upper Body Sports Model 2.0 and the general model?

A: It is aimed at the sports vertical field, emphasizing sports knowledge base, multimodal sports data integration and scenario-based applications (competitive training, event analysis, and national fitness), and is planned to be opened in phases, with the goal of directly reaching the national team and public training scenarios (October 25).

Q: What is the key value of the OneScience platform?

A: It provides a ready-to-use engineering foundation for large scientific models: supports domestic GPUs, pre-set multidisciplinary model templates, integrates computing power and data management, and significantly shortens the scientific research reproduction and online launch cycle (released from October 23 to 25).

Q: How does Mondelez implement the path of “reducing costs and increasing efficiency” by introducing generative AI?

A: It will first be used for high-frequency materials such as social media and e-commerce pictures, texts, and short videos. The short-term goal is to reduce costs by 30%-50%. In the medium term, it will be expanded to short TV advertisements, and will be launched as early as the holiday season in 2026, and the dual review of "human review and content standards" will be maintained (October 24-25).

Q: Why is the enthusiasm for AI spilling over into the energy and M&A markets?

A: AI data centers have high loads and are growing rapidly, which directly increases regional electricity and gas-fired power demand, driving up transactions in the US natural gas assets/pipeline/LNG industry chain; the policy rhythm of grid connection and capacity expansion has become the key to the implementation of the Compute project (October 24-25).

Q: How can developers and businesses avoid the risk of chatbots becoming “flattering”?

A: Introduce anti-"flattering" preferences during the training phase, improve fact-checking and citation transparency; strengthen safety switches and human review in sensitive scenarios such as medical/legal scenarios at the product level, and try to avoid replacing professional advice with "self-pleasing" feedback (Research and comments from October 24-25).

Q: Which links in the AI industry chain are investors more interested in this year?

A: In addition to computing power and GPUs, financial institutions emphasize cybersecurity, AIOps, data center infrastructure, and the practical application scenarios of "AI+industry"; at the same time, they use decentralization to cope with the uncertainty of valuation and realization (October 24).