The cover of "Nature" on November 21, 1985 is a "map" from China. This map is an illustration in "Haiguo Wenjianlu" compiled by Chen Lunjiong, the general soldier of the Qing Dynasty. It is a testimony to the Chinese people's beginning to recognize and explore the world. China".
At that time, China was in the wave of reform and opening up, and it seemed surprising and immature for Chinese technology to enter the international perspective.
In a blink of an eye, 38 years have passed, and with the rapid development of China's science and technology, more and more Chinese scientific researchers and scientific research institutions have been listed on "Nature", which is called the top of the scientific pyramid, as the first author.
On May 2, the main issue of "Nature" magazine published Baidu's breakthrough research results in the field of biocomputing, and published it as the fastest in the form of "Accelerated Article Preview".
In this article titled "Algorithm for Optimized mRNA Design Improves Stability and Immunogenicity", the mRNA sequence optimization algorithm LinearDesign is proposed. It not only created many possibilities for the integration and breakthrough of AI, biology, medical treatment and other fields, but also drew a new stroke for Chinese technology to go global.
In this paper, we can read content far beyond its achievements, such as the many possibilities of AI, the scientific responsibility of a technology company, and the great ambition of China's technology to go global.
"Unexpected" biological computing
I believe everyone knows that "Nature" only includes those researches that are breakthrough and significant in the field of basic science.
Perhaps it can be said that Baidu's exploration in the field of biological computing vividly tells what it means to "never underestimate the possibility of AI".
This matter starts with the new crown vaccine. Although the haze of the epidemic has quietly passed, this epidemic has made the world see the importance of vaccines, especially mRNA vaccines, in public health.
The so-called mRNA is a natural molecule that can produce target proteins or immunogens, thereby activating specific immune responses in the body to fight various pathogens. Moreover, it has important advantages such as mRNA does not contain viral components, has no risk of infection, and has a short research and development cycle. It is the killer weapon for humans to fight against the new crown epidemic.
However, in such an important field, there are still some problems in the development of vaccines and drugs, such as how to efficiently design mRNA sequences that are stable and more druggable?
In order to solve this problem, Baidu "unexpectedly" combined AI technology with vaccine research and development based on its accumulation in the AI field, and developed the LinearDesign algorithm that was published in the "Nature" magazine.
This algorithm uses Lattice Parsing technology in natural language processing to optimize the mRNA vaccine sequence, thereby improving the stability and effectiveness of the vaccine.
( American cardiologist and genomicist Eric J. Topol shared the Baidu LinearDesign algorithm on Twitter)
Here is an important point. The LinearDesign algorithm can be said to use the knowledge in the field of linguistics to overcome the problems in biomedicine. Although the two fields cannot be said to be irrelevant, they are basically separated by thousands of miles, but Baidu's exploration of AI technology has allowed the two to complete a marriage of thousands of miles, and the LinearDesign algorithm was born.
In May 2020, in the face of the raging epidemic, Baidu Research Institute launched the world's first mRNA vaccine gene sequence design algorithm LinearDesign, and announced that it will be open to vaccine research and development institutions and research centers around the world for free. LinearDesign can complete the mRNA vaccine sequence design of the new coronavirus in 16 minutes, which greatly accelerates the research and development efficiency of the new coronavirus vaccine.
Taking the Spike protein of the new coronavirus as an example, if you use the traditional method to find a stable mRNA sequence, you need to check 10632 mRNA sequences, which can be called an astronomical number, but with the LinearDesign algorithm, you can find the most stable candidate sequence within 11 minutes .
In December 2020, Baidu won the first "AI For Good (artificial intelligence for good)" award of the AIconics Awards held by The AI Summit, the world's top artificial intelligence summit, for its outstanding contributions of LinearFold and LinearDesign algorithms in the fight against the new crown epidemic.
In addition to the application value, the LinearDesign algorithm also demonstrates the new possibilities of AI as an underlying scientific exploration tool in the fields of biology and medicine. This kind of breakthrough across disciplines is more cherished by "Nature" magazine and even the global scientific community.
Unlocking the Depth and Breadth of AI
Why can Baidu make AI+biological computing research such as LinearDesign algorithm bigger and stronger? This may be another key piece of information we must read.
There is no other reason for it, it's just a lot of hard work. From 2012 when AI was in the ascendant to today's global popularity of AI, Baidu's persistence and exploration of AI technology has been consistent during this period. This persistence has permeated all aspects of AI technology, including infrastructure research, algorithm iteration, and interdisciplinary integration of AI. Many of these areas are full of unknowns and challenges, and do not conform to the traditional pursuit of short-term interests by enterprises. But building a solid infrastructure and exploring a forward-looking technology layout is the key to a company winning the future.
AI + biological computing, as the extensibility direction of AI technology, can better reflect Baidu's technical layout thinking of "downplaying short-term interests and focusing on long-term development". Biocomputing can solve key problems such as protein analysis and new drug research and development, and its value is huge. Taking new drug research and development as an example, this field has a "three 10" characteristics, that is, 10 years, 1 billion US dollars, and 100,000 people can develop a product. An effective new drug, and if AI is used as the drug development engine, it will greatly change the paradigm of new drug development and bring inestimable value.
Although this road is of great value, it is full of challenges, and it is difficult to achieve commercial returns in the short term. However, Baidu is still willing to rely on the advantages of AI technology and infrastructure to lay out this future road in advance. As early as 2018, Baidu officially launched research in the direction of computational biology. Over the past few years, Baidu has built a three-dimensional innovation map in this field that is surrounded by infrastructure, algorithms, and ecology, such as the LinearDesign algorithm just mentioned, and the LinearFold algorithm that can greatly speed up RNA structure prediction. A series of innovations have begun It outlines Baidu's differentiated advantages in the biological computing algorithm layer, and the biological computing platform-propeller PaddleHelix based on the flying paddle ecology has started the construction of AI+ biological computing underlying development tools, laying the foundation for the exploration of biological computing in the industry and academia. Facing the industrial ecology, Baidu provides the industry with pre-trained large models for the fields of compound molecules, protein molecules, and genomics information, and actively devotes its own technical advantages to industry-university-research collaborations. Just like at the level of AI infrastructure, Baidu emphasizes the parallelism of technology, ecology, and infrastructure. In the field of biocomputing, the same strategic implementation method also promotes the positive development of Baidu itself and the biocomputing industry.
Enterprises carry out interdisciplinary exploration and underlying technological innovation, emphasizing long-term and heavy accumulation. Only by building a complete technical sequence and technical system can they make a blockbuster on this basis. Whether it is in the field of AI+biological computing or more basic AI technology, long-termism has always been Baidu's trump card.
This logic of exchanging heavy accumulation for high efficiency is manifested in all aspects of Baidu's technological innovation. For example, within one month after the release of Wenxin Yiyan, four technical iterations were completed, which increased the reasoning efficiency by 10 times compared with the original version. Where does the efficiency of this technology iteration and industrial application come from? Among them, Flying Paddle supports the whole process of Wenxin Yiyan from development and training to inference deployment, and through joint optimization, it greatly improves the training and inference of large models. It can be seen from this that the long-term accumulation of underlying technologies and tool chains has laid the foundation for the efficient upgrading of new technologies.
Baidu is opening up the depth of AI technology in this way, and Chinese technology is also opening up the breadth of going to the world.
Big country science and technology go to the world
In November 1985, "Nature" magazine discussed China's scientific research level in high-energy physics, space exploration, earthquake prediction and other fields at that time, showing to the world a China with vigorous scientific research and innovation vitality.
In July 2008, "Nature" took the opportunity of the Beijing Olympics to release a special issue for China, introducing to the world that "the number of papers published in China is more than that of other countries except the United States", and confirmed the status and value of China's scientific research strength .
For decades, the "Chinese elements" in "Nature" have been perfectly synchronized with the development of China's science and technology. For example, in 1997, only 0.4% of the original scientific research papers published by "Nature" involved Chinese authors. In 2017, this has increased to about 15%. The change in this figure just proves the change of China's technology going global.
In this process, a role is indispensable, and that is the role played by technology companies in the overall scientific research environment. Since the information revolution, we can see that many major innovations are completed by enterprises in the technological landscape of developed countries in Europe and the United States. Taking AI+biological computing as an example, AlphaFold2 created by DeepMind has attracted wide attention. Alphabet, the parent company of Google, has continued to promote the implementation of this algorithm in the fields of medical health and new drug research and development. It has created ChatGPT's OpenAI, and is also exploring AI disease diagnosis and treatment, AI protein structure prediction and other fields.
Judging from the worldwide trend, in the important field of AI+biological computing, the main innovation positions should be in technology companies, and technology companies need to be at the core to form a benign ecology of industry-university-research interaction. The layout and aggressiveness of Baidu in the field of biocomputing complete this link in China's technological landscape.
China's science and technology going global, and moving towards self-reliance and self-improvement in science and technology, cannot do without the continuous upgrading of scientific research capabilities of enterprises. From this perspective, Baidu Biocomputing published a paper in Nature as the first completion unit, which can be seen as a signal of a new stage. In 1985, the world saw China's science and technology; in 2008, China's comprehensive scientific research capabilities ranked among the top in the world; today, Chinese scientists, academic institutions, and technology companies can all be listed in "Nature" as the first author, proving that China already has Diversified and three-dimensional technological capabilities.
The innovation and uniqueness of China's science and technology have extended from academia to industry, with a more global vision and more responsibility as a major country.
