These five AI developments will shape 2021 and beyond

The 12 months 2020 was profoundly difficult for residents, firms, and governments world wide. As covid-19 unfold, requiring far-reaching well being and security restrictions, synthetic intelligence (AI) functions performed an important function in saving lives and fostering financial resilience. Research and improvement (R&D) to boost core AI capabilities, from autonomous driving and pure language processing to quantum computing, continued unabated.  

Baidu was on the forefront of many vital AI breakthroughs in 2020. This article outlines five important advances with implications for combating covid-19 in addition to reworking the way forward for our economies and society.

1. AI and vaccine improvement

The pattern—and why it issues. It usually takes years, if not many years, to develop a brand new vaccine. But by March 2020, vaccine candidates to struggle covid-19 had been already present process human assessments, simply three months after the primary reported circumstances. The report pace of vaccine improvement was partly because of AI fashions that helped researchers analyze huge quantities of information about coronavirus.

There are tens of 1000’s of subcomponents to the outer proteins of a virus. Machine studying fashions can kind by means of this blizzard of information and predict which subcomponents are essentially the most immunogenic—i.e., able to producing an immune response—and thereby information researchers in designing focused vaccines. The use of AI in vaccine improvement might revolutionize the best way all vaccines are created sooner or later.

Baidu’s improvements. In February, Baidu opened its LinearFold AI algorithm for scientific and medical groups working to struggle the virus. LinearFold predicts the secondary construction of the ribonucleic acid (RNA) sequence of a virus—and does so considerably quicker than conventional RNA folding algorithms. LinearFold was in a position to predict the secondary construction of the SARS-CoV-2 RNA sequence in solely 27 seconds, 120 instances quicker than different strategies. This is important, as a result of the important thing breakthrough of covid-19 vaccines has been the event of messenger RNA (mRNA) vaccines. Instead of standard approaches, which insert a small portion of a virus to set off a human immune response, mRNA teaches cells the best way to make a protein that may immediate an immune response, which drastically shortens the time span concerned in improvement and approval. 

To assist mRNA vaccine improvement, Baidu later developed and launched an AI algorithm for optimizing mRNA sequence design referred to as LinearDesign, which goals to unravel the issue of unstable and unproductive mRNA sequences in candidate vaccines.  

In addition to opening up entry to LinearFold and LinearDesign for researchers world wide, Baidu additionally shaped a strategic partnership with the National Institute for Viral Disease Control and Prevention, a part of the Chinese Center for Disease Control and Prevention. Following an outbreak at Beijing’s Xinfadi market in June, Baidu’s AI expertise allowed authorities to finish genome sequencing of the coronavirus pressure inside 10 hours, serving to curb the outbreak. In December, Baidu unveiled PaddleHelix, a machine learning-based bio-computing framework aimed toward facilitating the event of vaccine design, drug discovery, and precision medication.

2. Fully automated driving and the rollout of robotaxis

The pattern—and why it issues. Autonomous driving expertise continued to mature in 2020, with the business’s main firms testing driverless vehicles and opening up robotaxi companies to the general public in varied cities. Fully automated driving, which allows rides and not using a human security driver on board, will be crucial for the scalability and commercialization of autonomous driving.

Baidu’s improvements. Over the previous 12 months, Baidu launched the Apollo Go Robotaxi service within the Chinese cities of Changsha, Cangzhou, and Beijing—together with in busy industrial areas—turning into the one firm in China to begin robotaxi trial operations in a number of cities.

These developments are a results of Baidu’s steady innovation in creating AI programs that may safely management a automobile in advanced highway circumstances and resolve the vast majority of attainable points on the highway, impartial of a human driver.

At Baidu World 2020, its annual expertise convention, Baidu additionally demonstrated its totally automated driving functionality—the place the AI system drives independently with out an in-vehicle security driver. To assist totally automated driving, Baidu developed the 5G Remote Driving Service, a security measure whereby distant human operators can take management of a automobile within the occasion of an distinctive emergency. Baidu’s achievement of totally automated driving, and the rollout of its robotaxis, suggests a constructive outlook for the commercialization of the expertise within the close to future.

Source: Baidu

3. Applied pure language processing

The pattern—and why it issues. In 2020, pure language programs turned considerably extra superior at processing facets of human language like sentiment and intent, producing language that aligns with human talking and writing patterns, and even visible understanding, that means the potential to precise understanding about a picture by means of language. These pure language fashions are powering extra correct search outcomes and extra subtle chatbots and digital assistants, main to higher person experiences and creating worth for companies.

Baidu’s improvements. Baidu launched a brand new multiflow sequence framework for language era referred to as ERNIE-GEN. By coaching the mannequin to foretell semantically full blocks of textual content, ERNIE-GEN performs at an elite degree throughout a spread of language era duties, together with dialogue engagement, query era, and abstractive summarization.

Baidu’s vision-language mannequin ERNIE-ViL additionally achieved important progress in visible understanding, rating first on the VCR leaderboard, a dataset of 290,000 questions constructed by the University of Washington and the Allen Institute for AI, that goals to check visible understanding means. ERNIE-ViL additionally achieved state-of-the-art efficiency on five vision-language downstream duties. Visual understanding lays the inspiration for pc programs to bodily work together in on a regular basis scenes, because it entails each understanding visible content material and expressing it by means of language. It will be essential for bettering the standard of human-machine interplay.

4. Quantum computing

The pattern—and why it issues. Quantum computing made important inroads in 2020, together with the Jiuzhang pc’s achievement of quantum supremacy. This carries significance for AI, since quantum computing has the potential to supercharge AI functions in comparison with binary-based classical computer systems. For instance, quantum computing could possibly be used to run a generative machine studying mannequin by means of a bigger dataset than a classical pc can course of, thus making the mannequin extra correct and helpful in real-world settings. Advanced applied sciences reminiscent of deep studying algorithms are additionally enjoying an more and more important function within the improvement of quantum computing analysis. 

Baidu’s improvements. Baidu achieved plenty of technical breakthroughs in 2020 that promise to bridge AI and quantum computing. In May, Baidu launched Paddle Quantum, a quantum machine studying improvement toolkit that may assist scientists and builders rapidly construct and prepare quantum neural community fashions and present superior quantum computing functions. The open-source toolkit each helps builders constructing quantum AI functions, and helps deep studying fans develop quantum computing. In September, Baidu entered cloud-based quantum computing with the launch of Quantum Leaf, which offers quantum improvement kits reminiscent of QCompute, and can shorten the life cycle of quantum programming and assist understand a ‘closed-loop’ quantum instrument chain.

Source: Baidu

5. AI chips

The pattern—and why it issues. AI {hardware} continued to develop in 2020, with the launch of a number of AI chips personalized for specialised duties. While an abnormal processor is able to supporting AI duties, AI-specific processors are modified with explicit programs that may optimize efficiency for duties like deep studying. As AI functions turn out to be extra widespread, any improve in efficiency or discount in value can unlock extra worth for firms that function a large community of information facilities for industrial cloud companies, and can facilitate the corporate’s inner operations.

Source: Baidu

Baidu’s improvements. At Baidu World 2020, the corporate provided a glimpse into its next-generation AI processor, the Kunlun 2, which it plans to place into mass manufacturing in early 2021. The chip makes use of 7 nanometer (nm) processing expertise and its most computational functionality is over 3 times that of the earlier era, the Kunlun 1. The Kunlun chips are characterised by excessive efficiency, low value, and excessive flexibility, which might assist a broad vary of AI functions and eventualities, serving to foster higher AI adoption and decreasing utilization prices. More than 20,000 Kunlun 1 chips have now been deployed to assist Baidu’s search engine and Baidu Cloud companions since they launched in 2018, empowering industrial manufacturing, good cities, good transportation, and different fields.

This content material was produced by Baidu. It was not written by MIT Technology Review’s editorial employees.

We will be happy to hear your thoughts

Leave a Reply

Reset Password