JD.com accelerates the race for artificial intelligence (AI) in China with the publication of its large linguistic model (LLM), even as regulatory authorities are preparing to introduce regulations to manage generative AI services.
JD.com , China’s second-largest online shopping platform, said that its LLM ChatRhino (or yanxi in Chinese) now supports several vertical sectors such as logistics, retail, healthcare, and finance. The model includes 70% general data and 30% “smart native” data, said the e-commerce player, which has a logistics branch and a business unit in the health field.
ChatRhino boasts a base of 100 billion parameters, compared to 10 billion parameters for its previous model, Vega, presented at the beginning of last year. Vega topped the list in the General Language Understanding Evaluation (GLUE) test, beating out the models from Microsoft and Facebook, said JD.com in a press release last Thursday.
An AI for health in “a few minutes”
GLUE measures and classifies natural language processing on the basis of nine tasks, covering a range of “data set sizes, text genres and degrees of difficulty”. OpenAI’s GPT-4 is said to have more than a trillion parameters based on eight models. But this information is not confirmed by the company. Its previous model GPT-3 has more than 175 billion parameters, while GPT-2 has 1.5 billion.
ChatRhino offers more than 100 training and inference optimization tools that, according to JD.com , support the development of domain-specific applications, allowing customers to create their own specialized models faster.
The provider claims that a generative AI model for the healthcare sector, for example, can be built in “a few minutes” with two algorithm engineers, while the traditional method usually requires a week and at least 10 scientists.
Immediate uses for online marketing
JD Health’s own extended language model, Jingyi Qianxun, is built on ChatRhino and has been trained on medical scenarios to automatically deploy services, including telemedicine.
Online merchants can also use ChatRhino to create a whole series of visuals, marketing posters and product images, from a single product image. The AI model can reduce the cost of producing each visual element by 90% and reduce the time required to complete the task from a week to half a day, according to JD.com.
Provisional laws to guide the deployments of generative AI
The launch of ChatRhino comes the same week that China is introducing provisional regulations to manage generative AI services in the country.
The new laws, which will come into force on August 15, are necessary to ensure the healthy development of technology and preserve national security and public interests, the Chinese government said.
In a joint statement issued by various agencies, including the Cyberspace Administration of China (CAC) and the Ministry of Science and Technology, the government notes that while generative AI creates new opportunities for economic and social development, it also brings challenges such as fake news and data security and privacy risks.
The beginning of coaching in China
The provisional legislation presents various measures aimed at facilitating a healthy development of technology, while protecting national and public interests as well as the rights of citizens and companies, according to the press release.
Generative AI developers, for example, will have to ensure that their pre-training and model optimization processes are carried out in compliance with the law. This includes using data from legitimate sources that respect intellectual property rights. If personal data is used, the consent of the individual must be obtained or this must be done in accordance with existing regulations.
Measures must also be taken to improve the quality of training data, in particular their accuracy, objectivity and diversity.
Generative AI providers assume legal responsibility for the information generated
Under the provisional laws, providers of generative AI services assume legal responsibility for the information generated and its security. They will have to sign service level agreements with the users of their services, thus clarifying the rights and obligations of each party.
When illegal content is discovered, the service provider must take various measures such as preventing its transmission and rectifying its use in model training. The competent authority must also be informed.
In addition, service providers will have to take the necessary measures if a user engages in illegal activities using the generative AI service. These measures include restriction functions, suspension or termination of the service, maintenance of relevant records and notification to the competent authority.
Fear of abuse
Service providers who violate the new laws are subject to sanctions under the various Chinese laws in force, including the network security law, the data security law and the personal information protection law. In cases where there are no provisions relating to violations, a warning will be issued, accompanied by a correction order that must be executed within a given period. Failure to comply with these orders may result in the suspension of services.
In April, China published a draft law, saying that the development of generative AI technologies such as ChatGPT could give rise to abuses if it were not regulated.
Separate legislation came into force in January. It establishes basic rules to prevent the misuse of “Deep synthesis” technologies, including “deepfakes” and virtual reality. Anyone using these services must label the images accordingly and refrain from using the technology for activities that violate local regulations.
Massive nationwide AI investment plan
In May, the Chinese government unveiled plans to build AI industrial centers and technology platforms throughout the country. To date, development plans have been launched for 18 national AI pilot zones and 32 innovation platforms, including in Beijing and Tianjin.
Besides JD.com , local players such as Tencent and Alibaba have also announced their efforts to offer or integrate generative AI into their products. Alibaba Cloud unveiled in April its AI platform called Tongyi Qianwen, which is currently available in China for beta testing and as an API for developers.
The Chinese cloud service provider has also launched a partnership program to stimulate the development of AI applications for vertical sectors, including finance and petrochemicals.
Source: “ZDNet.com “