UNESCO is scheduled to hold a Global Forum on the Ethics of Artificial Intelligence on the 5th and 6th, with important topics including equality and diversity related to artificial intelligence. Over the past year or so, countries have been racing to develop generative AI in their own languages, with deep considerations in the field of digital sovereignty behind them.
Multilingual products are emerging.
After ChatGPT, an English-language generative artificial intelligence developed by OpenAI in the United States, was launched at the end of 2022, researchers from many countries have followed up with the "large language model" technology, forming a boom in the development of local language generative AI in many places around the world.
In China, large language model products such as the company's "Wenxin Yiyan" and iFLYTEK's "Xinghuo" have shown good Chinese processing capabilities after rapid iteration, and have been used as productivity tools by many enterprises. Chen Xiaoping, director of the Working Committee on Artificial Intelligence Ethics and Governance of the Chinese Society of Artificial Intelligence, told Xinhua News Agency: "China's relevant research and development progress is relatively fast, some cross-domain large models continue to make new progress, and the development trend of vertical field large models is also very good." ”
In France, there were 79 generative AI startups as of September last year. Among them, Mistral Artificial Intelligence has a valuation of nearly 2 billion euros, becoming a leading company in the field of artificial intelligence in Europe. In December last year, the company released the "Mixtral 8x7b" model, which is available in five languages: French, Spanish, Italian, English, and German.
YandexGPT, developed by the Russian internet giant Yangdekos, successfully passed the literature subject in the first unified examination in October last year, 15 points higher than the minimum admission threshold for literature subjects in universities. Dmitry Masyuk, manager of the company's search and advertising technology team, said it has surpassed ChatGPT's 3Version 5.
In addition, Japan, South Korea, Vietnam and other countries have also developed generative AI in their own Chinese languages, and Singapore plans to develop large language models that can understand Indonesian, Malay and Thai.
Focus on "sovereign artificial intelligence".
It can be seen from the successive advent of products in different languages that the huge impact of generative AI is gradually emerging on a global scale, and the importance of "sovereign AI" is also becoming increasingly prominent in various countries.
"Sovereign AI" is a concept that has been frequently mentioned since the explosion of generative AI, which roughly refers to the country's support for the construction of AI infrastructure, the training of large models with data that is in line with the country's culture and way of thinking, and the support for the formation of ecosystems for local products.
France links AI technology to the country's industrial and digital sovereignty. In setting up the Commission on Generative AI last year, Minister of Economy, Finance and Industry, Digital Sovereignty, Bruno Le Maire, emphasized: "Thirty years ago, we missed the ...... of the digital revolutionWe can't afford to miss this second chance. In this national race, France has important resources and we must go faster and further. ”
India's Minister of State for Electronics and Information Technology, Rajiv Chandrasekhar, recently publicly stated: "We have made up our minds that we must have our own sovereign artificial intelligence. ”
Nvidia, an American chip company, said it would support countries in developing sovereign artificial intelligence. "People realize that they can't give their country's knowledge, their culture, to other people and then sell AI back to them. Nvidia CEO Jensen Huang recently talked about "sovereign artificial intelligence".
And this reflects a major problem in the development of sovereign artificial intelligence in various countries. At present, American companies are leading in the field of artificial intelligence, and if other countries want to keep up with the trend, they must adopt American products, which may lead to subsequent development being constrained by others. Not only hardware chips, but also in terms of software, many large models around the world are also "shelled" open source software in the United States. For example, the Japanese large model Swallow announced by the Tokyo Institute of Technology and other institutions in Japan in December last year is based on the LLAMA 2 model of the American "Yuan" company.
Mina Saideze, a German data analysis expert, said: "Most of the AI language models use English, and most of the best language models are from the United States......Germany and other European countries rely on American companies for many artificial intelligence technologies, and they are not independent in this field at all. ”
Be wary of the new digital divide.
The uneven global development of generative AI could lead to new inequalities. UN Secretary-General António Guterres said at Davos earlier this year: "We need to make systematic efforts to increase access to AI so that developing economies can benefit from its enormous potential." We need to bridge the digital divide, not deepen it. ”
Chen Xiaoping pointed out that a large number of data analysis shows that with the popularization and application of new technologies, the digital divide in some countries or regions is deepening, and many people are worried that similar consequences will occur between large models in different languages. For the new problems arising in the era of artificial intelligence, it is necessary to use new perspectives such as high-quality development to pursue the unity of economic and social benefits, so as to effectively ensure and improve the well-being of the whole society and all mankind.
In this process, China has shown good momentum in the field of artificial intelligence, which is expected by some developing countries. Ronny Lins, director of the Brazilian Center for Chinese Studies, told reporters that the development of artificial intelligence in Brazil is not sufficient, while China and the United States have made a lot of investment in the field of artificial intelligence. One of the global problems that AI can cause is the impact on employment. It is expected that China will play to its strengths and contribute to the creation of new jobs for young people through the use of new technologies.
The Global Forum on the Ethics of Artificial Intelligence encompasses all of these topics. At a time when AI is competing in different languages, Gabriela Ramos, UNESCO's Assistant Director-General for Social and Human Sciences, said: "We need truly global coordination and knowledge sharing to build a responsible AI ecosystem that benefits everyone. ”
Reporter Huang Kun Participating reporters: Xu Yongchun, Chen Chang, Qian Zheng, Sun Yi, Du Zheyu, Bian Zhuodan, Feng Yasong, Lu Rui, Li Gong).
Xinhua News Agency, Beijing, Feb. 5.