24-Hour AI News: Sugon OneScience and Genos Launch; Australia Launches Compliance Inquiry on AI Chatbots

In the past 24 hours (October 23rd), China focused on "computing network + industry model + application implementation" (Soongon's OneScience platform, genomic foundational models, AI conversational assistants, and the "Millisecond Computing" initiative), while North America and Europe focused on "regulation and industrial capital" and "underlying capabilities" (Australia's eSafety legal notice, the UK's official announcement of OpenAI's UK industrialization and data hosting, Redwood's financing and introduction of NVIDIA, Google's quantum algorithm, and S&P's launch of the AI Enhanced Index). This briefing only includes events that have not been reported by this channel in the past 72 hours or have seen substantive new developments.

1. Sugon releases China's first "one-stop development platform for large scientific models" OneScience (China)

1. Released at the China Computer Conference on October 23, it is positioned as an integrated platform for "scientific large-scale model research and development, training, evaluation, and deployment."
2. The model has been reproduced and optimized on domestic GPUs, and the model code is available for download to the industry.
3. Aims to shorten the cycle from code to application of scientific research models and improve the connection between scientific research computing power and open source ecology.

2. Genos, the world's first deployable "universal genome base model" with tens of billions of parameters, was released (China)

1. Released on October 23, with a parameter scale of approximately 10 billion, it supports "million base pairs" ultra-long context and "single base resolution" recognition.
2. Integrate 636 high-quality human genomes at the "telomere-to-telomere (T2T)" level as training data to more comprehensively cover population differences.
3. Adopting a “mixture of experts (MoE)” architecture to reduce inference costs, aiming at the implementation of precision medicine and population genetics research.

3. Quark launches its "Bean Bag-like" AI conversational assistant, deeply integrating the Qwen closed-source model (China)

1. On October 23, the new dialogue assistant was launched, emphasizing the integrated experience of "AI search + dialogue".
2. Introducing Tongyi Qianwen's latest closed-source model capabilities to complement complex queries and multi-round question-and-answer scenarios.
3. Expand the AI reach of Alibaba apps for high-frequency application scenarios such as education and efficiency.

4. MIIT promotes the "Millisecond Computing in Metropolitan Areas" campaign (China)

1. Disclosed on October 23: By 2027, a "millisecond-level" computing network capability system with full coverage, high efficiency and smooth operation in the metropolitan area will be basically formed.
2. Data side: my country has 10.85 million standard racks of computing facilities in use, and the scale of intelligent computing power is about 788 EFLOPS.
3. Significance to the industry: Provides network and computing power guarantees for low-latency scenarios such as AI reasoning, robot and vehicle-road collaboration in industrial sites.

5. China Mobile and the Chinese Academy of Sciences signed an agreement on "Research on Key Technologies of Computing Networks for General Artificial Intelligence" (China)

1. It was disclosed on October 23 that central enterprises and research institutes are jointly tackling computing network technology for general artificial intelligence.
2. Objective: Collaboratively promote systematic innovation in computing power, algorithms, and data to support the evolution of intelligent computing networks.
3. Provide industry-university-research collaboration for the subsequent supply-side reform of "millisecond computing" and wide-area intelligent computing.

6. Australia eSafety issues "child safety" legal notices to four AI chat platforms (Australia)

1. On October 23, Character.ai, Glimpse.AI (Nomi), Chai, and Chub.ai were asked to explain their measures for protecting minors and managing content.
2. Failure to report as required may result in a fine of up to US$825,000 per day, enforced in accordance with the Online Safety Act.
3. Release regulatory signals: Generative "chat-type" applications must meet higher compliance transparency and security thresholds.

7. UK announces OpenAI expansion of data hosting and industry collaboration in the UK (UK)

1. On October 23, the British government announced that OpenAI will provide local data hosting options for British companies to strengthen "sovereign data" and compliance capabilities.
2. With the advancement of AI applications in the public sector, it is claimed that 240,000 "working days" of administrative time can be saved.
3. Collaborate with recent policy tools such as “AI Growth Labs/Growth Zones” to create a “local AI” ecosystem.

8. Redwood Materials secures $350 million in funding, with NVIDIA participating (USA)

1. On October 23, battery recycling unicorn Redwood completed a new round of financing to expand its recycling-material supply chain capabilities.
2. Investors include NVIDIA, pointing to a long-term layout for the coupling of "energy, computing power, and supply chain in the AI era."
3. Release side signals: Upstream energy and materials have become the key "second curve" of AI infrastructure.

9. Google discloses “milestone-level” progress in quantum algorithm research (USA)

1. It was announced on October 22nd, claiming new progress in quantum algorithms.
2. Although it is still far from the general generative AI application chain, it is expected to lay the foundation for the long-term "AI × quantum" computing paradigm.
3. Industry focus: How to transform into engineering capabilities and software stack ecosystem for actual quantum acceleration.

10. S&P Dow Jones Indices to Launch First “AI-Enhanced” Industry Rotation Index (North America)

1. Released on October 23, it systematically introduced predictive modeling methods into index construction for the first time.
2. As a methodological innovation, it may promote the compliance of the “AI×Index Engineering”.
3. Impact on institutions and asset management: The toolbox will be expanded, but attention should be paid to the stability of the samples and the risk of information leakage.

Frequently Asked Questions (Q&A)

Q: What is the key leap of the Genos model compared to previous genome models?

A: The Genos model released on October 23 has approximately 10 billion parameters, supports million-base pair context and single-base resolution recognition, and its training data covers 636 T2T-level human genomes. It also adopts the MoE architecture to control inference costs, making it more usable and fair for precision medicine.

Q: What direct changes will “millisecond computing” bring to enterprises?

A: By 2027, we aim to achieve "millisecond-level" computing scheduling capabilities in the metropolitan area. Combined with the current intelligent computing power of approximately 788 EFLOPS and the scale of 10.85 million standard racks, scenarios such as industrial quality inspection, robot collaboration, vehicle-road collaboration, and edge inference will receive lower latency and more stable computing network services.

Q: How binding is the “legal notice” issued by Australia’s eSafety?

A: According to the "Online Safety Law", the inquiry obligations and compliance requirements for chat-type generative AI platforms are clarified. Violations or non-cooperation may be fined up to US$825,000 per day, and will be implemented in conjunction with the "Basic Online Safety Expectations".

Q: What can companies gain from the UK’s announcement of OpenAI’s local data hosting?

A: For B-side users operating in the UK, they can choose to store their business data within the UK to reduce cross-border compliance and sovereign data risks, which will help accelerate the implementation of AI in public services and regulated industries.

Q: What is the difference between S&P’s “AI Enhanced Index” and ordinary quantitative index?

A: The index systematically introduces predictive modeling (AI) as an optional factor within the industry rotation framework, which is an upgrade of the index engineering methodology; however, the index is still a long-term tool, and attention should be paid to the consistency between backtesting and actual trading.