On July 4, it was reported that at the 2024 World Artificial Intelligence Conference, China Mobile announced that it would collaborate with partners such as China National Petroleum Corporation (CNPC), Huawei, and iFlytek to promote the widespread application of multimodal big models in the oil industry, new energy, and other fields. With "AI Empowerment, Industrial Innovation," CNPC has made significant progress in the field of artificial intelligence, including big models.

In May of this year, CNPC finalized its partnership and signed cooperation agreements with China Mobile, Huawei, and iFlytek, officially launching the construction of the Kunlun big model collaboration. Another big model for the energy and chemical industry led by a central enterprise will emerge.

▲Previously, CNPC's Dai Houliang met with Baidu's Robin Li and iFlytek's Liu Qingfeng. Source: Baidu and iFlytek official WeChat accounts

CNPC's implementation of multiple big models reflects the urgency of digital transformation in the oil and gas sector. Interestingly, in its multiple forays into big models, CNPC has chosen different "partners," reflecting its cautious and discerning approach to technology layout while also hinting at the competitiveness among technology giants.

Embracing Big Models Proactively

Large models such as ChatGPT and Sora have created a global sensation, and CNPC has proactively embraced big models, paving a new path for digital transformation in the energy industry.

The 2024 government work report mentioned deepening the research, development, and application of big data, artificial intelligence, and other technologies, carrying out the "AI+" initiative, and building a digital industry cluster with international competitiveness.

Shortly before the two sessions, the State-owned Assets Supervision and Administration Commission of the State Council held a special promotion meeting on artificial intelligence for central enterprises with the theme of "AI Empowerment, Industrial Innovation." The meeting proposed that central enterprises should plan the development of artificial intelligence in their overall work, deeply promote industrial innovation, and accelerate the layout and development of the AI industry.

CNPC has previously built multiple big models, which directly confirms that CNPC's exploration and investment in artificial intelligence have always been at the forefront of the energy and chemical industry.

As early as 2003, CNPC began developing the ultra-large seismic data processing and interpretation integrated software system Geoeast. By 2023, the version has evolved from the initial 1.0 to 4.0, forming a geophysical exploration software covering multiple scenarios from land to sea. This software has been widely used in over 70 enterprises such as Sinopec and CNOOC, and has been adopted by more than 30 universities as teaching software and research platforms, becoming one of the three mainstream geophysical exploration software globally.

In 2022, the Geophysical Technology Research Center of BGP, a subsidiary of CNPC, developed a new intelligent velocity pick-up technology called Smarter-VeloPick. This technology has improved the efficiency of velocity spectrum interpretation in GeoEast software by 6 to 20 times.

In January 2024, the PetroChina Research Institute of Petroleum Exploration and Development released PetroAI, an independently developed AI large language model. This is the earliest, largest, and most functional large language model in the oil and gas exploration and development field.

In February 2024, a joint venture established by the China Petroleum Pipeline Bureau Design Institute and Baidu released WisGPT, the first domestic AI big model in the field of oil and gas storage and transportation. It can achieve human-computer interaction through text, voice, and other forms, providing intelligent support for enterprises in digital management, as well as oil and gas storage and transportation engineering survey, design, construction, and operation.

In addition, CNPC, which is planning for digital transformation, has also ventured into the "Internet+" field in the downstream sales segment early on. In 2015, after Sinopec successively collaborated with Internet giants such as Tencent and Alibaba, CNPC also followed suit, officially partnering with Tencent to enter the era of "Internet + oil and gas".

Drawing on Experience

In the domestic market, there are already many domestic big model products in the first tier, such as iFlytek Spark, Baidu Ernie Bot, and Alibaba Pangu. In selecting partners for the Kunlun big model, CNPC ultimately chose iFlytek after negotiations and other selection methods.

Before CNPC finalized its partnership for big model construction, there were undercurrents of AI competition among many technology giants. For example, Baidu's official WeChat account showed that on May 7, Robin Li led the Baidu AI team to visit the headquarters of CNPC and held talks with Dai Houliang, Chairman and Party Secretary of China National Petroleum Corporation. A week later, on May 14, iFlytek's official WeChat account reported that Liu Qingfeng led iFlytek executives to visit the headquarters of CNPC and also held talks with Chairman Dai Houliang.

It may be slightly surprising to outsiders that Baidu did not appear on the list of companies participating in the construction of the big model. Baidu and the China Petroleum Pipeline Bureau Design Institute not only established a joint venture in April 2023 but also released a big model product in February 2024. However, the existing cooperation foundation failed to give Baidu an advantage in larger-scale and higher-level cooperation with CNPC.

After years of exploration and testing by its various branches, CNPC has formed a relatively clear plan for big models. According to Dai Houliang, CNPC plans to build the Kunlun big model to reflect the technical characteristics of big models and accelerate the upgrading of the innovation chain throughout the production and operation process. In selecting partners, CNPC has gone through rigorous selection, and ultimately, China Mobile with advanced middleware equipment, Huawei with superior computing power, and iFlytek with obvious advantages in general big model foundations have emerged.

It is understood that iFlytek not only has the collaborative support of China Mobile and Huawei but also has deeper technological accumulation in the field of enterprise AI and in big model development capabilities and experience. The release of iFlytek Spark V3.5 has formed a complete solution for the cloud implementation of big models and has been implemented in energy and other industry scenarios. In addition, iFlytek Spark is currently the only national-produced computing power-based big model, which has a natural advantage in autonomy and controllability.

For the oil and gas industry, the "AI+" track is full of opportunities. According to data from the CCID Research Institute of the Ministry of Industry and Information Technology, in 2023, the adoption rate of generative AI in China reached 15%, with a market size of approximately 14.4 trillion yuan. It is expected that by 2035, China's generative AI economic scale is expected to exceed 30 trillion yuan. Big models can support all AI tasks and are representative of the new generation of AI, also known as general AI. They can not only significantly lower the threshold for developing AI products but also improve the performance and versatility of AI, with very broad prospects. CNPC's proactive layout in AI and parallel deployment of big models are likely part of this strategy.

In recent years, in the downstream sector of the oil and gas industry, scenarios such as refueling and non-oil business applications at gas stations have become increasingly "sci-fi," with technologies such as AR and robots enabling car owners to refuel while sitting in their cars, and automatic car washes, automatic fuel dispensers, and smart logistics making gas stations more intelligent.

The deployment of big models in upstream production areas will bring disruptive digital transformation to the energy industry. Big models can handle and understand more complex and variable business scenarios such as image recognition, voice and video processing, and language applications. If applied to the oil sector, they can enhance the business processing capabilities of oil industry professionals. Perhaps in the near future, drilling platforms, oil pipelines, and gas stations can be remotely monitored, and oil workers can sense operating parameters such as pressure, temperature, sound, and inventory at any time with a simple finger movement.