news

한어Русский языкEnglishFrançaisIndonesianSanskrit日本語DeutschPortuguêsΕλληνικάespañolItalianoSuomalainenLatina

on august 28, a press conference on the construction of the 33 billion parameter kunlun big model with the theme of "ai navigation, intelligent energy" was held in beijing. the china petroleum kunlun big model was officially released at the conference, and the latest achievements of the big model construction, such as the application of seismic interpretation big model, logging processing interpretation big model, employee assistant application, and industry application, were demonstrated.

zhou song, chief accountant and member of the party leadership group of china national petroleum corporation, said that china national petroleum corporation is fully committed to promoting the innovative development of big models, focusing on building a large-scale application ecosystem, strengthening close cooperation with china mobile, huawei, and iflytek, and continuously iterating to form industry big models with larger parameter scales, higher-quality data sets, and more powerful ai middle platforms, creating higher value and more "artificial intelligence + energy and chemical" landmark achievements represented by the kunlun big model.

it is reported that petrochina held a seminar on the construction of kunlun big model on may 28, and signed a cooperation agreement on the construction of kunlun big model with china mobile, huawei, and iflytek, officially launching the construction of kunlun big model. at present, the project has preliminarily completed the top-level design, and in accordance with the "five ones" action plan, trained 8 big models, developed 18 application scenarios, built a rich industry and professional data set, completed the customized design of ai middle platform, and built a smart computing environment. it has initially formed a methodology for the implementation of big models in large enterprises, achieved phased results, and become the first industry big model in the energy and chemical industry that has passed the national filing.

xu jiajia, vice president of iflytek and president of antelope, said in his speech that in order to support the research and development of kunlun big model, iflytek and antelope invested a lot of core backbone forces in terms of manpower, and nearly 100 people participated in this project. in terms of large language model training, while fully supporting the training work required by the filing, we actively responded to the overall planning needs of the group and completed the training of 13 billion and 33 billion parameter industry big models as scheduled; in terms of scenario development, we completed the research and development and debugging of 8 scenarios. it was a challenging task to quickly complete this series of work in three months, but with the joint efforts of the digital intelligence institute, china mobile, huawei and other teams, we achieved technological breakthroughs and progress.

in three months, iflytek worked closely with teams from petrochina, china mobile, huawei and other organizations to create the kunlun big model. with the support of the kunlun big model's powerful understanding and generation capabilities, learning and reasoning capabilities, and multimodal fusion capabilities, the team completed the development and debugging of eight scenarios, including oil knowledge q&a, refining knowledge q&a, industry information, industry bidding announcements, meeting assistants, official document writing, rules and regulations q&a, and intelligent operation q&a, to help employees improve their work efficiency and quality.

empowering the industry is an important mission of kunlun big model. based on more than half a century of industry data accumulation by petrochina, the project training has created an ai search and question-answering engine in the field of energy and chemical industry. on this basis, the industry application has been developed. it will become an expert in the energy and chemical industry around everyone, making the acquisition of industry knowledge more convenient, more accurate and more efficient. in the era of intelligence, the employee assistant application driven by kunlun big model will cover all scenarios and intelligent office services, and gradually become a reliable, caring and efficient "smart assistant" for every employee.

with the rapid development of artificial intelligence technology, big models are being used more and more widely in energy and chemical industries such as oil and gas. the in-depth combination of application scenarios such as smart stations, "one-stop" oil field services, smart pipelines, intelligent reservoir geological research, and smart well sites with artificial intelligence big models has brought revolutionary changes to many aspects of oil and gas exploration, development, and production.

on june 27, iflytek spark big model v4.0 and related landing applications were released, comprehensively improving the seven core capabilities of the big model base. it not only ranked first in 8 international mainstream test sets, leading domestic big models, but also achieved overall transcendence of gpt-4 turbo in text generation, language understanding, knowledge question and answer, logical reasoning, and mathematical ability.

in order to help energy renewal, on the same day, antelope used the general capabilities of iflytek spark big model as the core technical foundation, combined with the actual needs of the energy industry, to create a solution with three major features: cloud-edge collaboration, autonomous control, and data and model security, covering six major industry scenarios: wind, light, water, fire, nuclear, and storage.

during the oil and gas exploration stage, the energy big model can analyze the collected geological data and improve the accuracy and efficiency of oil and gas exploration through multimodal information fusion technology; during the oil and gas development stage, it can predict the production performance of oil wells and provide a scientific basis for production increase measures; in the oil and gas production and operation links, the application of the energy big model covers equipment fault diagnosis, production parameter optimization, safety monitoring and other aspects.

xu jiajia proposed three proposals. the first is to continue to climb the height of large model technology. the research and development of large model technology is a systematic and continuous work. iflytek and antelope will continue to develop new general artificial intelligence technologies, optimize model structures, improve model performance, and inject continuous vitality into the kunlun large model.

the second is to promote the depth of integration of ai and scenarios vertically. china national petroleum corporation has profound accumulation in the entire industrial chain of the energy and chemical industry. in the next step, we will continue to focus on the characteristics of the energy and chemical industry, combine application scenarios such as "ai+oil and gas", "ai+refining", and "ai+manufacturing", and further promote artificial intelligence technology to meet the needs of industrial scenarios and improve the level of intelligence of business processes.

the third is to let oil workers feel the warmth of the human-machine collaborative model. with the advancement of artificial intelligence technology, the perceptual intelligence of machines will gradually surpass that of humans and assist humans in the industry. through human-machine coupling, machines are equipped with "brains" to free employees from complicated work and focus on more creative work. in the future, oil workers can realize oil extraction in suits and create more digital iron men "wang jinxi". (reporter wang qiao)