Breaking News

RealAI receives another 10 million financing: Tsinghua AI Research Institute incubated with a total amount of 100 million

RealAI receives another 10 million financing: Tsinghua AI Research Institute incubated with a total amount of 100 million

[This article is from the TMTpost feature column “Fast Company”]

Recently, RealAI, an AI technology and industry solutions provider, has completed an angel + round of financing of tens of millions of RMB. This round of financing was led by Songhe Capital and followed by Tongdu Capital. It will be mainly used for team expansion, R&D infrastructure construction and marketing, etc.

Relais Wisdom was established in July 2018 and incubated from the Artificial Intelligence Research Institute of Tsinghua University. The company is developing unsupervised learning methods and interpretable algorithms to solve the problem of missing labeled data in the industry, provide safe and controllable AI technology solutions, and help people make better decisions with the help of algorithms.

In January last year, Rayleigh Wisdom completed an angel round of financing of tens of millions, which was jointly led by BV Baidu Ventures and Zhongke Chuangxing, and followed by the Shuimu Tsinghua Alumni Fund.

So far, within one year, RealAI has received a total of nearly 100 million yuan in financing. Behind this, more is the capital market’s recognition of RealAI’s technical achievements, talent team, and business model.

RealAI receives another 10 million financing: Tsinghua AI Research Institute incubated with a total amount of 100 million

Develop the third-generation artificial intelligence learning algorithm

Currently, machine learning tasks include supervised learning, unsupervised learning and reinforcement learning. Among them, supervised learning refers to deriving prediction functions from labeled training data. To put it simply, it is to perform algorithm analysis and predict labels through given data.

At present, the entire market environment is still dominated by the second-generation AI technology represented by deep neural networks. However, the second generation of artificial intelligence is mainly based on supervised learning trained with large amounts of data. The technical route seriously lacks generalization ability and lacks interpretability. It is easy to be deceived and give wrong judgments, and the results are difficult to be trusted.

The third-generation artificial intelligence algorithm researched by RealAI is implemented through unsupervised or semi-supervised learning – inferring conclusions from unlabeled training data, and finding hidden scopes by given small sample data. Using this AI algorithm, even if it is not clear what the output value is, it can finally calculate its trend range.

With the substantial increase in data volume and computing power, the current artificial intelligence landing scenarios are extending from limited fields such as speech recognition and face recognition to higher-value application fields such as financial decision-making, industrial production operations, medical diagnosis, and autonomous driving. Behind this, higher requirements will be placed on the security and reliability of artificial intelligence, which also means a new round of AI algorithm iteration and the upgrade of the AI ​​industry model.

RealAI provides a feasible path for the development of the next generation of reliable, trustworthy and secure artificial intelligence with the dual-driven model of “data + knowledge”.

In terms of team, RealAI has strong scientific research strength. RealAI was incubated in the AI ​​Research Institute of Tsinghua University. The CEO, Dr. Tian Tian, ​​graduated from the Department of Computer Science of Tsinghua University. During the 2019 World Artificial Intelligence Conference (WAIC), he was awarded the title of “Young AI Scientist” by the AI ​​Young Scientists Alliance. Under the age of elite list.

Tian Tian, ​​CEO of RealAI

Academician Zhang Yan, academician of the Chinese Academy of Sciences and dean of the Institute of Artificial Intelligence of Tsinghua University, and Professor Zhu Jun, director of the Basic Theory Research Center of the Institute of Artificial Intelligence of Tsinghua University, jointly serve as the chief scientists of the company. The AI ​​Research Institute team of Tsinghua University led by Academician Zhang Cymbal has been researching and exploring new technologies such as Bayesian deep learning, interpretable learning, and AI against attack and defense.

In addition, the core backbone of the team comes from well-known companies such as Alibaba and Tencent, and most of them have backgrounds in prestigious schools such as Tsinghua University.

“It is generally judged in the industry that in the field of the third generation of artificial intelligence based on logical cognition, there will be companies with a market value of more than 100 billion US dollars in China.” In Songhe Capital’s view, RealAI has become an absolute leader in this track.

As an old shareholder, Mi Lei, the founding partner and co-CEO of Zhongke Chuangxing, said in an interview with Titanium Media that the third-generation artificial intelligence algorithm technology developed by RealAI is a trend in the future, especially in the field of Industrial Internet of Things. This algorithm will play a great role in practical applications.

Mi Lei told Titanium Media that he believes that the future industrial Internet of Things will not be the era of large sample data. Algorithms will become an indispensable technical means for the future industrial Internet of Things, with huge commercial value.

Three major directions of layout, so that technology can be commercialized

At present, RealAI has reached cooperation with a number of enterprises using the third-generation artificial intelligence technology, and its core customers currently cover large financial institutions, industry leading industrial enterprises, large central enterprises, etc., helping enterprises to improve full-stack business AI capabilities and improve operational efficiency. Lower overall costs. RealAI will mainly land in the three major fields of AI finance, AI industry and AI security, and its layout will enable the technology to be commercialized.

In the field of AI finance, RealAI proposes that the third-generation artificial intelligence has better model performance in scenarios such as no label, less label, cold start and uncertainty processing, and has the advantage of interpretability from black box to white box. Through the end-to-end machine learning modeling platform RealBox, it takes advantage of its self-developed algorithms. RealAI helps financial institutions operate long-tail assets more precisely in the field of big data risk control, recover and reject customers at zero cost, and obtain more interpretable prediction results. Realize the intelligent improvement of business and improve the reliability of the whole life cycle of financial scenarios. At present, RealAI has cooperated with a number of leading banks.

For the AI ​​industrial field, RealAI is committed to the optimization of equipment asset operation and production cost and efficiency optimization. In the manufacturing process, intelligent visual inspection technology is used to improve the stability of quality inspection and reduce labor costs; in the operation process, AI algorithms are used to optimize the system to improve efficiency; through the analysis of equipment monitoring signals to find out abnormalities, so as to realize the health status of equipment assessment and prediction to improve service life. The PV module quality inspection system provided by RealAI has significantly improved the production efficiency after the actual deployment of the production line, and the analysis and prediction accuracy of sensor monitoring signals using Bayesian deep learning is also significantly better than the traditional expert model.

AI security is a new business track opened up by RealAI. Aiming at the potential loopholes and security risks existing in the widely used AI algorithms such as face recognition and target detection, RealAI mines new algorithm loopholes based on the world’s leading adversarial attack and defense technology, and uses adversarial training and other methods to further build an AI firewall to achieve detection and defense. The dual purpose of attack behavior enhances the security of existing algorithms, and can be applied to security protection in the fields of face-scanning systems and smart security.

In addition, for deepfake content such as Deepfake and ZAO, RealAI has developed more accurate detection products, which are combined with traditional content auditing scenarios to create a healthier and safer content environment. Specific product functions include Deepfake deepfake video detection, new identification Review of pornographic/violent terrorist content, identification of counterfeit and shoddy products, etc.

The Links:   SKKH162-08E NL6448AC63-01