Everybody had a tough time in 2020, from individuals to businesses to governments. Artificial intelligence (AI) solutions were critical in preventing human suffering and bolstering economic stability as covid-19 spread and imposed severe health and safety constraints. R&D efforts to improve foundational aspects of AI, such as autonomous driving, natural language processing, and quantum computing, persisted.
Many of the most significant advances in artificial intelligence in 2020 were pioneered by Baidu. This article summarises five major developments that could change the future of our economy and societies and have an impact on the fight against covid-19.
A.I. and the creation of vaccines
The current trend, and why it’s significant. Creating a brand new vaccination might take a long time, sometimes decades. However, by March of 2020, only three months after the first reported cases, human testing were being conducted on vaccine candidates to combat covid-19. Using AI models, scientists were able to quickly sift through mountains of coronavirus data, which accelerated vaccine development to a previously unheard-of pace.
A virus’ exterior proteins consist of tens of thousands of smaller components. Researchers can use machine learning models to sift through this mountain of information and determine which subcomponents are the most immunogenic (able to elicit an immune response). The use of AI to the vaccine industry has the potential to radically alter the future of vaccine production.
New developments from Baidu. In an effort to aid medical and scientific organisations combating the virus, Baidu made available in February its LinearFold artificial intelligence technology. LinearFold is an algorithm that predicts the secondary structure of a ribonucleic acid (RNA) sequence at a much faster rate than previous RNA folding algorithms, making it ideal for use in virus research. LinearFold was 120 times faster than previous approaches at predicting the secondary structure of the SARS-CoV-2 RNA sequence in under 27 seconds. This is noteworthy since the discovery of messenger RNA (mRNA) vaccines has been the most significant advance in the field of covid-19 vaccines. In contrast to traditional methods, which entail inserting a small part of a virus to elicit a human immune response, mRNA teaches cells how to manufacture a protein that can elicit an immune response, drastically reducing the amount of time needed for development and approval.
Later, Baidu launched LinearDesign, an artificial intelligence system for optimising mRNA sequence design, with the goal of addressing the issue of unstable and ineffective mRNA sequences in vaccine candidates.
Baidu has announced a strategic agreement with the National Institute for Viral Disease Control and Prevention, which is a branch of the Chinese Center for Disease Control and Prevention. This partnership will allow researchers all around the world to have access to LinearFold and LinearDesign. After a coronavirus epidemic at Beijing’s Xinfadi market in June, authorities were able to sequence the entire genome of the virus strain within 10 hours thanks to AI technology developed by Baidu. Baidu introduced PaddleHelix in December; it is a bio-computing platform that uses machine learning to speed up the process of creating vaccines, finding new drugs, and developing precision medicine.
The introduction of robotaxis and the advent of fully autonomous vehicles.
The current trend, and why it’s significant. In 2020, the leading businesses in the autonomous driving industry continued to refine their technology, conducting tests on driverless cars and launching robotaxi services in a number of cities. Scalability and commercialization of autonomous driving will require fully automated driving, which allows rides without a human safety driver on board.
New developments from Baidu. Baidu is the only Chinese business to begin trial operations of robotaxis in various cities, having done so over the previous year with the launch of the Apollo Go Robotaxi service in Changsha, Cangzhou, and Beijing (particularly in busy commercial areas).
These advancements are the result of Baidu’s unceasing research and development of artificial intelligence (AI) systems capable of safely guiding a car in a wide variety of road conditions and solving the vast majority of potential problems without the need for a human driver.
Baidu also showcased its completely automated driving capacity at its annual technology conference, Baidu World 2020. In this mode, the AI system drives autonomously without an in-vehicle safety driver. Baidu created the 5G Remote Driving Service, a safety feature that allows human operators to take control of a car remotely in the event of a critical incident, to facilitate completely automated driving. Full automation of driving is now a reality, thanks in large part to Baidu’s efforts, and the company’s debut of robotaxis bodes well for the eventual widespread implementation of this technology.
Natural language processing (NLP) techniques used in real-world situations; data from Baidu 3.
The current trend, and why it’s significant. Specifically, by the year 2020, natural language systems had significantly improved their ability to process sentiment and intent in human language, generate language that aligns with human speaking and writing patterns, and even achieve visual understanding, which is the ability to express understanding about an image through language. Better search results, as well as more complex chatbots and virtual assistants, are all powered by natural language models, which is improving the user experience and providing value to businesses.
New developments from Baidu. In order to facilitate the creation of new languages, Baidu has introduced a new multiflow sequence architecture called ERNIE-GEN. ERNIE-GEN excels at a wide variety of language generation tasks, such as dialogue engagement, question generation, and abstractive summarization, since it is trained to predict semantically full blocks of text.
The Vision-Language Model Used by Baidu Significant advancements were made in visual understanding by ERNIE-ViL as well; it came in first place on the VCR leaderboard, which uses a dataset of 290,000 questions developed by the University of Washington and the Allen Institute for AI to assess visual understanding skills. On five vision-language downstream tasks, ERNIE-ViL also outperformed state-of-the-art systems. Knowledge visual content and being able to express that understanding through language forms the basis for computer systems to physically engage in everyday scenarios. This will be essential for advancing the state of the art in machine learning and human-machine interaction.