Image source: Generated by Unbounded AI

In the rapidly developing wave of science and technology, AI has transformed from a science fiction concept into a driving force in reality.

From affecting lifestyles to reshaping work styles; from improving production efficiency to optimizing decision-making processes; from reshaping traditional industries to spawning new business models, AI is no longer limited to conversation and chat, but Become an important force in promoting the development of new productive forces.

The new era driven by AI is quietly unfolding.

AI-driven, from technology to productivity

In the past year, AI has been shining brightly. Neuralink successfully completed the first human brain-computer interface implantation surgery; the video generation model Sora and the multi-modal model GPT-4o promoted a leap in generative AI technology; at the Nobel Prize Ceremony, AI was crowned for the first time...

From daily life to the forefront of science, AI is outlining a world where technology and humans are deeply intertwined.

Recently, Canalys, a world-renowned independent analysis agency for the technology market, released data stating that in the third quarter of 2024, cloud infrastructure service spending in mainland China reached US$10.2 billion, a year-on-year increase of 11%, returning to double-digit growth. .

This growth data means that the practical application of AI is becoming a key driving force for the development of cloud computing. To simply understand, the increase in the application of cloud computing power is due to the continuous expansion of the use of large models, and the growth of large model applications reflects the in-depth implementation of AI technology in actual scenarios.

That is exactly what happened. The 2024 "Global Digital Economy White Paper" pointed out that the number of large AI models in the world has reached 1,328, of which China accounts for 36%, ranking second in the world.

It is worth noting that China’s proportion of actual AI applications ranks first in the world. McKinsey data shows that 46% of users in China often use AI in work and life, higher than 39% in North America and 33% in Europe. This data not only reflects the high acceptance of AI technology by Chinese users, but also confirms China’s significant advantages in the implementation of AI scenarios.

Currently, China is embarking on a differentiated AI development path with its broad application scenarios and industrial chain integration capabilities.

From manufacturing to medical care, from traffic management to energy optimization, China’s AI development path is deeply rooted in the needs of the real economy. AI is no longer just a cold technology in the "virtual world", but an infrastructure for efficient decision-making and productivity improvement in the real economy.

This transformation not only demonstrates the practicality and plasticity of AI, but also marks the arrival of the era of "new quality productivity" driven by AI.

The core of new productivity is to reshape production relations and improve industrial efficiency through the combination of new technologies, new models, and new applications. In this major transformation, AI is not only the representative of new productivity itself, but also the core driving force that empowers other fields to achieve productivity upgrades.

AtAt the 2024 Central Economic Work Conference, "new quality productivity" was emphasized many times.

The meeting clearly stated that “we must coordinate the relationship between cultivating new kinetic energy and updating old kinetic energy, and develop new productive forces in accordance with local conditions.” It is necessary to “lead the development of new productive forces with scientific and technological innovation and build a modern industrial system.” "Carry out the 'artificial intelligence +' action and cultivate future industries."

Today, AI is not only a key component of new productivity, but also a force that promotes industrial intelligence. It profoundly affects the development direction of the industry. Through intelligent transformation, it helps traditional industries transform from "inefficient" to "inefficient". "Efficient", the leap from "artificial" to "intelligent".

Thousands of industries and industries outside the dialog box

Recently, "Global Times", China Association for Science and Technology, and Tsinghua University jointly released the "New Quality Productivity" "Industrial Practice Demonstration Cases" covers five popular fields: artificial intelligence, advanced manufacturing, new energy, biomedicine, and low-altitude economy. Five scientific and technological breakthrough cases have been launched in each field.

With careful observation, it is not difficult to find that AI is playing an increasingly important role in many industries, driving the transformation and upgrading of the industry.

Take the field of biomedicine as an example. The "Double Ten Law" is a long-term problem in this field, that is, the drug research and development cycle lasts for more than 10 years and costs more than 1 billion US dollars. Among them, the traditional Chinese medicine industry due to its The complexity of multiple components, multiple targets, and multiple diseases often faces innovation bottlenecks.

However, the application of AI in the medical field may be expected to break through this law and greatly accelerate the process of drug research and development.

For example, the "Digital Materia Medica" large model created by the biomedical company Tasly has accurately tuned and learned more than 40 million documents through the fusion training of the Huawei Cloud Pangu large model. , prescription optimization has significant effects. At the same time, relying on intelligent question answering and precise computing capabilities, the "Digital Materia Medica" large model is also promoting the standardization and intelligence of traditional Chinese medicine research and development.

Yunnan Baiyao, a century-old enterprise, has also developed the "Lei Gong Model" based on the ancient model, covering scenarios such as the popularization of traditional Chinese medicine knowledge, assisted decision-making, and digital marketing, and solved the problem of "planting good seeds" in the traditional Chinese medicine industry chain. medicine, sell good medicine, talk about good medicine, prescribe good medicine” and other key issues.

At the bottom of the acceleration of large models to cultivate new productivity for thousands of industries, computing power is the key to building large model application capabilities, and it is also the basis for realizing the widespread application of AI in various industries.

For example, in the logistics industry, the vertical large-scale language model "Fengyu" launched by SF Technology is based on the surging computing power provided by Shengteng AI cloud service, achieving efficient development and resource utilization of AI applications. . The application of this technology has significantly improved SF Express’s operational efficiency and customer experience.

iFlytek also teamed up with Ascend Cloud to create China’s first ultra-large-scale domestic computing platform - the "Flying Star One" large model platform. Currently, the Spark model is completely developed and trained based on Shengteng computing power.It has been iterated to the V4.0 Turbo version and has become the only large open model based on national computing power training.

Thanks to its complete AI infrastructure, Huawei Cloud has explored a wide range of AI implementation scenarios. The "China Industry Large Model Market Report, 2024" by Sullivan, an international authoritative organization, shows that Huawei Cloud ranks first in the market share of government affairs and industrial finance in the industry large model field, and ranks among the leaders in medical, pharmaceutical, meteorological and automotive industries. quadrant.

In the future, AI for Science

The potential of AI goes far beyond industry applications. It may also be used in science A paradigm revolution has begun in the field.

Philosopher of science Thomas Kuhn proposed in "The Structure of Scientific Revolutions" that every era has a "paradigm" that dominates the development of science, but the transformation of paradigms requires huge efforts and costs.

From the perspective of the evolution of scientific research, humans have experienced the "empirical paradigm" based on observation, the "theoretical paradigm" of mathematical abstraction, the "computing paradigm" relying on computing tools, as well as machine learning and big data-driven "data paradigm".

Today, AI for Science (AI4S artificial intelligence-driven scientific research) is emerging and is regarded as the upcoming fifth paradigm. By deeply integrating data and physical models, AI4S breaks the bottleneck of traditional scientific methods on high-dimensional complex problems and provides a new path for scientific exploration.

“The scientific research paradigm of AI for Science breaks the boundaries of disciplines. It also solves the problem of previous scientific research, from topic selection to results, where one person or a few people in a small circle work together to tackle the problem, and the production of 'small workshops' is low and efficient. difficult problems, allowing the workshop model to transform into a platform scientific research model," E Weinan, an academician of the Chinese Academy of Sciences, commented.

Although AI4S is still in the preliminary exploration stage, its potential in multiple fields is gradually being recognized by the scientific community. This can be seen from the list of Nobel Prize winners in 2024.

On October 9, 2024, the Nobel Prize in Chemistry was awarded to Demis Hassabis, John Jumper and David Baker for their contributions to protein structure prediction and computational protein design.

This achievement relies on AlphaFold2 developed by DeepMind. This tool uses multiple sequence alignment data and uses a data-driven approach to successfully predict hundreds of millions of protein structures in less than 3 years. Its accuracy is close to the experimental level, expanding humankind's accumulation in the past half century by thousands of times.

Similar changes are emerging in other fields.

In the field of mathematics, AlphaGeometry has solved 83% of international mathematics Olympiads through a neural-symbolic hybrid system.geometry problem; in the field of physics, FermiNet improves the computational efficiency of multi-electron quantum systems through parameterized quantum wave functions, capturing 97% of the relevant energy in complex molecular systems; in the field of chemistry, LapNet breaks through variational Monte Carlo The computing bottleneck has greatly improved the applicability...

It can be said that AI4S is relying on these practical results to promote the further development of the technological revolution and new productivity.

As a practitioner of technological innovation, Huawei deeply understands the key to innovative research and development, and therefore has a better understanding of the scenario needs of scientific research institutions. Huawei Cloud has also taken the lead in the field of AI4S and has taken the lead in putting it into practice.

Through the Pangu model, Huawei Cloud empowers scientific research and solves many problems in the fields of drug research and development, genetic research, weather prediction, and agricultural breeding, such as accelerating drug research and development, supporting protein structure prediction, and promoting global Intelligent weather forecasts and more.

In addition, in order to support the one-stop development of scientific research institutions in the long term, Huawei Cloud has also built an AI4S platform based on Ascend Cloud computing power. The platform services have covered biomedicine, computational chemistry, earth science, quantum mechanics, etc. 10 Multiple fields and 80+ models have been engineered and productized, supporting out-of-the-box use and secondary development, helping agile innovation and improving users' scientific computing AI development efficiency.

Based on the surging computing power provided by Shengteng AI cloud service, the platform realizes efficient operation of the entire process from model training to inference, providing scientists with a flexible and efficient R&D environment.

In the future, the potential of AI4S will be released at a deeper level. Whether it is basic theoretical research on quantum computing or high-precision simulation of complex fluid simulations, continuous investment in AI4S is not only for technological breakthroughs, but also for the reshaping of scientific research paradigms.

“The new paradigm of AI-driven scientific research marks that the accelerator button for human exploration of unknown territories has been pressed.”

Professor and doctoral supervisor at the School of Computer Science and Technology, Fudan University, Shanghai Data Science Xiao Yanghua, director of the Key Laboratory, said that fundamentally speaking, the essence of human beings may lie in their transcendence. Today, human beings have used their own intelligence as a template to create AI, assisting human beings in improving their ability to understand and transform the world, and continue to write "creative innovations". World myth".

When we look back at the history of science, every paradigm change opens a new door to human cognition. When AI becomes an indispensable "collaborator" in a team of scientists, we may also usher in a new explosion of knowledge and technological leaps.

The future is already here, but the story of AI has just begun.