Jiang Ning, CTO of Immediate Consumer Finance: Will release a large model that combines generative and discriminative models
“Currently, generative AI has brought a lot of impact to financial institutions. The company will increase its layout in three capabilities in the field of artificial intelligence. The first is the decision-making intelligence of data, which is also the true vitality of large models; the second is Multi-modal language, text and video artificial intelligence, and the third one is real-time human-computer collaboration.” On the afternoon of July 12, Jiang Ning, CTO of Immediate Consumption, gave a speech on “The Road to Digital Transformation of Finance” at the 2023 Beijing News Shell Finance Summer Summit. stated on the forum. He also revealed that the large model to be released for immediate consumption is a combination of these three aspects.
Spend now Jiang Ning: A large financial model will be released in the near future, which will combine generative models and discriminative models
In fact, whether it is combined AI capabilities, continuous learning capabilities, or decision-making capabilities, they provide a better foundation for the wider application of large-scale data. He believes that data decision-making is the most critical capability of financial institutions, which can ensure that artificial intelligence is more stable and continuously assists in major decisions. This is why Immediate Consumption deploys these three core capabilities.
Since 2020, Beijing News Shell Finance has held three consecutive summer summits, each holding more than 10 special forums focusing on the annual economic theme. The 2023 Beijing News Shell Finance Summer Summit will focus on 10 major themes such as artificial intelligence, specialization, specialization, and ESG, and will come from national ministries and commissions such as the Ministry of Industry and Information Technology, the National Development and Reform Commission, the Ministry of Finance, the Ministry of Ecology and Environment, and the Ministry of Commerce, universities, and scientific research institutes. Research institutions such as institutes, social groups such as associations and societies, local government departments, and entrepreneurs from different industries conducted a collision of ideas.
Finance has always been the bloodline of the real economy and an important force supporting the high-quality development of the real economy. The arrival of the digital economy has accelerated the integration of artificial intelligence and finance, effectively promoting the digital transformation of finance. New generation information technologies such as artificial intelligence have also become a powerful thrust and new engine for finance to better serve the real economy.
Jiang Ning, CTO of Immediate Consumption, and many senior practitioners discussed “how AI can help the financial industry serve the real economy.”
“The current hot generative AI technology will have great potential.” Jiang Ning said that the advancement of generative AI lies in its strong transfer learning ability and generalization ability. However, most financial institutions currently use industrial-grade discriminant models, that is, certain To give results, unlike GPT, the answer it gives you may be wrong, but after training, the model accuracy will gradually improve. Therefore, the main idea for exploring large models in the future should be to effectively combine the generalization ability and transfer learning ability of generative models with discriminative models to solve practical problems of financial institutions.
As for possible problems in the process of commercial implementation, Jiang Ning believes that exploration should mainly focus on the four levels of “algorithm, computing power, data, and specifications”.
He explained that in terms of algorithms, as the environment changes, artificial intelligence capabilities can change, such as forming a broad ecology, and continuous feedback data from ecological participants makes large models more and more mature, rather than one-time models; in In terms of computing power, large models have higher requirements for the infrastructure of computing power, and a single computing power needs to be increased to seven or eight thousand; in terms of data, in order to make the model more efficient, it requires the participation of the whole society and financial institutions. However, due to the Restrictions such as technology and compliance are difficult to share among the industry and need to be explored; in terms of regulatory governance, most decisions in finance are very rigorous and the results must be stable. How to combine the decision-making models of financial institutions with the general methods of large models. Therefore, combined AI is needed, which can not only exert the generalization ability of large models, but also the combination ability of small models, so as to make decisions stable, safe and comprehensive. regulation.
The latest data shows that since its establishment in 2015, Immediate Consumption has more than 3,000 employees, more than 2,000 of whom are professionals in artificial intelligence, big data and other fields, and has 10,000 servers, 2,000 models and millions of Intelligent data decision-making system.
