On June 28, Hundsun Electronics and its subsidiary Hundsun Juyuan officially released a new digital financial product based on big language model technology, financial smart assistant Photon and a new upgraded smart investment research platform WarrenQ. In addition, LightGPT, a large model of Hang Seng Electronics' financial industry, also made its debut and announced the latest research and development progress.
Liu Shufeng, chairman of Hundsun Electronics, said that the large model is the latest breakthrough in information technology, which has refreshed people's understanding of machine intelligence and refreshed the industry's application of traditional AI models. In the era of large models, "data + algorithms + computing power" constitute the basic elements of the new paradigm.
Large-scale models have brought changes such as "voice-controlled everything" and "big talents" in commercial applications, but for the vertical field of finance, there are still limitations in terms of timeliness of domain knowledge, data security and privacy protection, and specific application support. . The financial industry needs to build a large-scale industry model to meet the specific needs of the vertical field, promote upstream and downstream cooperation in the large-scale model industry, and provide strong technical support for financial digital intelligence.
If financial institutions want to deploy and apply large models, they need to start with three steps: model selection, scale estimation, and scenario building. Today, as a technology company serving the financial industry, Hundsun Electronics has combined its own technical capabilities and in-depth understanding of financial business to create a large model of the financial industry and new digital intelligence products based on the large model, providing a new paradigm for the application of large models in the financial industry.
WarrenQ , the way to open a new generation of investment research
WarrenQ is a professional integrated investment research tool platform launched by Hang Seng Juyuan for investment research and investment scenarios. It empowers the entire process of "searching, reading, computing, and writing" investment research scenarios to improve investment research efficiency. At the same time, it is an integrated collaboration platform , breaking the traditional island of investment research information, and effectively assisting the digital ecological construction of investment research.
At the meeting, Bai Xue, deputy general manager of Hang Seng Juyuan, released the newly upgraded WarrenQ based on large-scale model technology. According to the actual needs of investment research scenarios for data analysis and intelligent processing, WarrenQ launched two AI tool products - WarrenQ-Chat and ChatMiner, which upgraded the previous "search, read, calculate and write" to "Chat, read, calculate and write", through intelligent dialogue way to help analysts and investment managers improve work efficiency.
WarrenQ-Chat is a chat product in the financial vertical field. It uses large-scale model superposition search and Juyuan financial database to easily obtain financial market conditions, information and data through dialogue instructions, and each generated dialogue supports original text traceability to ensure that the news The source can be traced, and financial professional reports can also be generated, easily realizing "voice control".
ChatMiner is a financial document miner based on a large model and vector database. It can quickly interpret specified documents according to user dialogue instructions, provide accurate retrieval and positioning, extract key information, and effectively integrate and summarize information. Refine or expand, intelligently process massive text data.
Bai Xue said that the new generation of WarrenQ opened up the investment research work scene through natural language interaction, and opened a new generation of intelligent investment research product form integrating "big model + data + tools". In the future, with the further maturity of the large-scale model of the financial industry, WarrenQ will also combine more scenarios to output intelligent tools to help the digital intelligence development of investment research.
Photon, financial intelligent assistant in the era of big models
With the maturity and application of large-scale model technology, automation and intelligence are gradually becoming a part of financial business processes. However, currently limited by the lack of data training and learning in the financial vertical field and the demand for data protection in the financial field, the general large model cannot provide comprehensive, accurate and instant answers in Chinese financial scenarios. Based on the actual needs of the large model of financial business applications, Hundsun Electronics took the lead in launching the financial intelligent assistant Photon, which solved the problems of technical connection, application connection, and real-time data security and compliance in the process of implementing the large model, and realized the intelligent upgrade and integration of the business system. refactor.
Le Shifei, CTO of Hundsun Electronics, said at the release that in the technical ecology layout of Hundsun's large-scale model products, Photon is an intelligent application service that connects "universal tool chain + financial plug-in tools + financial data + financial business scenarios" in series. Based on the LightGPT capability of the large model of the financial industry, Photonics can inject AI capabilities into financial institutions' investment advisory, customer service, operation, compliance, investment research, trading and other business systems, and become an AI assistant for financial practitioners.
The press conference showcased Photon's capabilities in consulting, creation, compliance, and operational scenarios. In the investment advisory business, Photon will become a consultant and content creation assistant. Investment consultants can input instructions, analyze customer intentions in real time through photon, accurately identify customer needs, and generate professional consulting opinions and suggestions based on user data. You can also use photon creation platform to automatically generate professional opinions, marketing copywriting, and portfolio analysis reports and other copywriting, making investment advisory content creation easier and more efficient.
In the compliance scenario, Photon will become the smart steward of the compliance system. Photon automatically generates compliance settings with one click according to compliance terms, and provides consultation questions and answers on investment compliance requirements. In the operation scenario, Photon is the intelligent helper of the operation system. Photon uses its powerful language processing capabilities to play a role in parameter extraction and intelligent analysis of non-standard documents, automatically screening documents that match the valuation system or TA system, and intelligently extracting key parameters.
On the basis of providing "model + tools", Photon gathers all kinds of financial data, including Juyuan financial database (markets, announcements, research reports, etc.), data generated by application systems, characteristic data, etc., and uses unified "standards" "Processing, storage, opening and use. At the same time, Photon has realized data desensitization, compliance, authorization and traceability in terms of data protection, access control, legal compliance, and operation auditing.
According to Le Zhifei, Hundsun Electronics will officially open its photonic capabilities to the public in September for trial use.
LightGPT , a large model of the financial industry
Due to its particularity, the financial industry has higher requirements for training data quality, model generation effect, and data security compliance. At this press conference, LightGPT, a large model of the financial industry created by Hundsun Electronics, also made its debut. It is understood that, as a professional large-scale model for the financial industry, LightGPT is more professional, more compliant, and lighter than general-purpose large models.
Bai Shuo, dean of Hang Seng Research Institute and chief scientist of Hang Seng Electronics, said that LightGPT has more professional financial corpus accumulation and processing and a more efficient and stable large-scale model training method, using more than 400 billion tokens of financial field data (including information, announcements, research newspapers, structured data, etc.) and language-enhanced data of more than 40 billion tokens (including financial textbooks, financial encyclopedias, government reports, regulations, etc.), and use them as the secondary pre-training corpus of large models, supporting more than 80+ financial The fine-tuning of exclusive task instructions enables LightGPT to have the ability to accurately understand the financial field.
At present, LightGPT has performed well in the evaluation of financial large model capabilities, including financial professional question and answer, logical reasoning, super long text processing ability, multi-modal interaction ability, code ability, etc., and ensures the compliance and security of content and instructions , is at the leading level in the industry, and can provide underlying AI capability support for financial business scenarios such as investment advisory, customer service, investment research, operation, risk control, compliance, and R&D.
In addition, LightGPT has more abundant and lightweight deployment methods, supports privatization/cloud deployment and flexible API calls, and the inference end only needs to be deployed with one machine and two cards. Financial institutions can customize and fine-tune large models based on LightGPT through private domain task data to create institutional-specific large models to meet individual needs.
According to Bai Shuo, LightGPT will complete a new round of financial capability upgrades at the end of September and officially open the trial interface.
At the meeting site, Hundsun Electronics also joined hands with industry ecological partners such as China Academy of Information and Communications Technology, Guotai Junan Securities, Shenwan Hongyuan Securities, Galaxy Securities, Industrial Securities, Great Wall Securities, E Fund Management, etc., to hold the "Hang Seng Digital Intelligence Financial New Product Co-construction Launching Ceremony". In the future, Hundsun Electronics will also actively participate in the domestic artificial intelligence ecological construction, and jointly lead the upgrading and transformation of financial digital intelligence with financial institutions and ecological partners, so as to help the high-quality development of the financial industry.