January 28, 2021

The five AI developments will shape the year 2021 and beyond

The five AI developments will shape the year 2021 and beyond.

Despite the deceptions of 2020, AI has accelerated its progress. Baidu increased its performance through vaccines, autonomous vehicles, language processing and quantum computing.

Citizens, businesses and governments around the world faced a profound challenge in the year 2020. As covid-19, which required far-reaching health and safety restrictions, expanded, artificial intelligence (AI) applications played a crucial role in saving lives and building economic resilience. Research and development (R&D) activities to improve basic AI capabilities, from autonomous driving and natural language processing to quantum computing, continued unabated.

In 2020, Baidu was at the forefront of many important AI developments. This article describes five significant developments with implications for the fight against covid-19, as well as for the transformation of the future of our economies and society.

  1. AI and vaccine development

Fashion - and why it matters. It usually takes years, if not decades, to develop a new vaccine. But by March 2020, covid-19 vaccine candidates were already being tested in humans, just three months after the first reported cases. The record speed of vaccine development was due in part to AI models that helped researchers analyze large amounts of data on the coronavirus.

There are tens of thousands of sub-components of the external proteins of a virus. Self-learning models can bypass this maelstrom of data and predict which sub-components are the most immunogenic - that is, capable of producing an immune response - and thus guide researchers in designing specific vaccines. The use of AI in vaccine development may revolutionize the way all vaccines are created in the future.

Baidu’s innovation. In February, Baidu opened its LinearFold AI algorithm to scientific and medical teams working on fighting the virus. LinearFold predicts the secondary structure of a virus’s ribonucleic acid (RNA) sequence, and does so significantly faster than traditional RNA folding algorithms. LinearFold was able to predict the secondary structure of the SARS-CoV-2 RNA sequence in only 27 seconds, 120 times faster than other methods. This is significant, because the key breakthrough for covid-19 vaccines has been the development of messenger RNA (mRNA) vaccines. Instead of conventional methods, which insert a small portion of a virus to trigger a human immune response, mRNA teaches cells to make a protein that can trigger an immune response, greatly shortening the time needed for its development and approval.

To support mRNA vaccine development, Baidu developed and subsequently published an AI algorithm to optimize mRNA sequence design, called LinearDesign, which aims to solve the problem of unstable and unproductive mRNA sequences in candidate vaccines.

In addition to opening access to LinearFold and LinearDesign to researchers worldwide, Baidu also formed a strategic partnership with the National Institute for Viral Disease Control and Prevention, which is part of the Chinese Center for Disease Control and Prevention. Following an outbreak in Beijing’s Xinfadi Market in June, Baidu’s AI technology enabled authorities to complete the genome sequencing of the coronavirus strain within 10 hours, helping to curb the outbreak. In December, Baidu introduced PaddleHelix, a machine-learning-based bioinformatics framework aimed at facilitating vaccine design development, drug discovery and precision medicine.

  1. Fully automated driving and deployment of robotaxis

The trend… and why it matters. Autonomous driving technology continued to mature in 2020, with the leading companies in the sector testing driverless cars and opening robotaxi services to the public in several cities. Scalability and commercialization of autonomous driving will require fully automated driving, allowing for travel without a human safety driver on board.

Baidu’s innovation. Last year, Baidu launched the Apollo Go robotaxi service in the Chinese cities of Changsha, Cangzhou and Beijing - including the busiest shopping areas - making it the only company in China to start robotaxi testing operations in several cities.

This development is the result of Baidu’s continuous innovation in perfecting AI systems that can safely control a vehicle in complex road conditions and solve most possible problems on the road, independent of a human driver.

At the annual Baidu World 2020 technology conference, it also demonstrated its fully automated driving capability, in which the AI system is driven independently without a safety driver on board the vehicle. To support fully automated driving, Baidu developed the 5G Remote Driving Service, a safety measure by which remote human operators can take control of a vehicle in the event of an exceptional emergency. Baidu’s achievement of fully automated driving, and the deployment of its robotaxis, suggests a positive outlook for the commercialization of the technology in the near future.

  1. Applied natural language processing

Natural language systems became significantly more advanced in processing aspects of human language such as feeling and intent, generating language that aligns with human speech and writing patterns, and even visual understanding, meaning the ability to express understanding about an image through language. These natural language models are driving more accurate search results and more sophisticated chatbots and virtual assistants, leading to a better user experience and creating value for businesses.

Baidu’s innovation. The company launched a new multiflow sequence framework for language generation called ERNIE-GEN. By training the model to predict semantically complete blocks of text, ERNIE-GEN performs at a level of excellence in a range of language generation tasks, including dialogue participation, question generation, and abstract summarization.

In addition, Baidu’s vision-language model, ERNIE-ViL, has made significant progress in visual comprehension, taking first place in the video-tape recorder rating table, a 290,000 question dataset built by the University of Washington and the Allen Institute for AI to test visual comprehension ability. ERNIE-ViL also achieved state-of-the-art performance on five subsequent vision-language tasks. Visual comprehension lays the foundation for computer systems to physically interact in everyday scenes, as it involves both understanding visual content and expressing it through language. It will be crucial to improve the quality of human-machine interaction.

  1. Quantum computing

Natural language systems became significantly more advanced in processing aspects of human language such as feeling and intent, generating language that aligns with human speech and writing patterns, and even visual understanding, meaning the ability to express understanding about an image through language. These natural language models are driving more accurate search results and more sophisticated chatbots and virtual assistants, improving the user experience and creating value for businesses.

Baidu’s innovation. The company launched a new multiflow sequence framework for language generation called ERNIE-GEN. By training the model to predict semantically complete blocks of text, ERNIE-GEN performs at a level of excellence on a range of language-generation tasks, including dialogue participation, question generation, and abstract summarization.

In addition, Baidu’s vision-language model, ERNIE-ViL, has made significant progress in visual comprehension, taking first place in the video-tape recorder rating table, a 290,000 question dataset built by the University of Washington and the Allen Institute for AI to test visual comprehension ability. ERNIE-ViL also achieved state-of-the-art performance on five subsequent vision-language tasks. Visual comprehension lays the foundation for computer systems to physically interact in everyday scenes, as it involves both understanding visual content and expressing it through language. It will be crucial to improve the quality of human-machine interaction.

  1. Quantum computing

The trend - and why it matters. Quantum computing made important advances in 2020, including the achievement of quantum supremacy for the Jiuzhang computer. This is important for AI, since quantum computing has the potential to overload AI applications compared to classic binary-based computers. For example, quantum computing could be used to run a learning model of generative machines through a larger data set than a classical computer can process, thus making the model more accurate and useful in real world environments. Advanced technologies, such as deep learning algorithms, are also playing an increasingly critical role in the development of quantum computing research.

Baidu innovations. Baidu achieved a series of technical advances in 2020 that promise to bridge the gap between AI and quantum computing. Last May, Baidu launched Paddle Quantum, a set of quantum machine learning development tools that can help scientists and developers quickly build and train quantum neural network models and provide advanced quantum computing applications. The open source toolkit supports developers building quantum AI applications and helps deep learning enthusiasts develop quantum computing. In September, Baidu entered cloud-based quantum computing with the launch of Quantum Leaf, which provides quantum development kits such as QCompute, and can shorten the life cycle of quantum programming and help realize a “closed loop” quantum tool chain.

  1. AI chips

AI hardware continued to be developed in 2020, with the release of several custom AI chips for specialized tasks. While an ordinary processor is capable of supporting AI tasks, specific AI processors are modified with particular systems that can optimize performance for tasks such as deep learning. As AI applications become more widespread, any increase in performance or reduction in costs can unlock more value for enterprises operating a large data center network for business services in the cloud, and can facilitate internal enterprise operations.

Baidu’s innovation. At Baidu World 2020, the company offered a look at its next-generation AI processor, the Kunlun 2, which it plans to put into mass production in early 2021. The chip uses 7 nanometer (nm) processing technology and has a maximum computational capacity more than three times that of the previous generation, Kunlun 1. Kunlun chips feature high performance, low cost, and high flexibility, which can be compatible with a wide range of AI applications and scenarios, helping to encourage greater AI adoption and reduce usage costs. More than 20,000 Kunlun 1 chips have already been deployed to support the Baidu search engine and Baidu Cloud partners since its launch in 2018, leveraging industrial manufacturing, smart cities, intelligent transportation and other fields.

Source: MIT Technology Review