Beijing has taken steps to regulate China's chatbots five months after ChatGPT sparked an artificial intelligence investment frenzy. This is a sign of the government’s determination to maintain tight regulatory control on technology that could define a new era.
This month, the Cyberspace Administration of China released draft rules for what is called generative artificial intelligent -- software systems like ChatGPT that can generate text and images in response to user questions and prompts.
Companies must adhere to the strict censorship regulations of the Chinese Communist Party, and websites and apps are prohibited from publishing materials that discredit China's leaders, or repeat forbidden history. Content of A.I. The content of A.I.
The companies will have to ensure that their chatbots are honest and do not violate intellectual property. They will also be required to register the algorithms behind their chatbots with regulators.
Experts said that although the rules aren't final and may be modified by regulators, engineers in China who were building artificial intelligence products had already begun to figure out how to integrate them into their products.
Chatbots have stunned governments around the globe with results that range from alarming and benign to A.I. generated. Artificial intelligence was used to create a fake picture of Pope Francis wearing a puffy jacket and to help students ace their college exams.
ChatGPT developed by OpenAI in the U.S., which has received $13 billion dollars from Microsoft, has prompted Silicon Valley to use the technology behind it to develop new products, such as video games and advertising. Sequoia Capital, a venture capital firm, estimates that A.I. Businesses could produce "trillions" of dollars in economic value.
Investors and entrepreneurs in China are racing to catch-up. The shares of Chinese artificial-intelligence firms have surged. Some of China's largest tech companies have made splashy announcements, including the ecommerce giant Alibaba, SenseTime which produces facial recognition software, and the search engine Baidu. Two start-ups that are developing Chinese alternatives to OpenAI technology have raised millions in funding.
ChatGPT does not exist in China. Faced with an increasing number of domestic alternatives, China quickly revealed its red lines in artificial intelligence. This was ahead of other countries who are still deciding how to regulate chatbots.
Kendra Schaefer is the head of technology policy at Trivium China in Beijing.
She added, "Because there is no two-party system, where both sides are arguing, they can simply say, "OK, we know that we need to do it, and we will revise later."
Developers struggle with the sometimes surprising and inaccurate results that chatbots can produce. China's rules appear to require a technical level of control over chatbots, which Chinese tech companies do not have. Microsoft, for example, is still working to improve their chatbots in order to eliminate harmful responses. China has much higher standards, which is the reason why some chatbots are already shut down or only available to a small number of users.
The experts are divided over how difficult it will become to train A.I. Systems must be consistent in their factual information. Some companies are unable to account for Chinese censorship laws, which can be very broad, constantly changing, and require the censorship even of specific dates and words, such as June 4, 1989 the day of Tiananmen massacre. Some believe that with time and enough work the machines will align themselves with truths and values, including political systems.
Analysts anticipate that the rules will be changed after consulting with China's technology companies. Regulators may soften the enforcement of rules so that they don't completely undermine the development of technology.
China has a history of censoring internet. In the 2000s, China built the most powerful dragnet of information on the internet. It intimidated Western companies such as Google and Facebook that were not compliant. It employed millions of workers to monitor online activity.
China's tech firms, which were required to follow the rules, thrived, defying Western critics, who had predicted that political control could undermine growth and innovation. Companies helped the government harness technologies like facial recognition and mobile phone to create a surveillance society.
The current A.I. The current A.I. He is also a fellow of the Carnegie Endowment for International Peace.
Researchers call this hallucination. Researchers call this hallucination. It runs contrary to the party's obsession about controlling what is said on social media, Sheehan explained.
He added that "Generative Artificial Intelligence put two of the main goals of the Party into tension: the control over information and the leadership in artificial intelligent."
Experts said that China's new rules are not just about politics. They aim, for example, to protect the privacy of individuals and their intellectual property as well as that of data creators. Models are being trained, which is a subject of concern worldwide.
Getty Images, a company that creates image databases, sued Stable Diffusion in February for using 12 million watermarked images to train its artificial intelligence system. Getty claimed this lowered the value of their images.
China has a larger push to answer legal questions regarding A.I. The use by companies of the underlying content and data. Beijing, as part a major institution overhaul, established the National Data Bureau in March to better define ownership, buying and selling data. The government body will also help companies build the data sets needed to train these models.
They are now deciding which data belongs to property and who is allowed to control and use it," explained Ms. Schaefer. She has published extensively on China's A.I. The initiative was described as "transformative" by the regulations.
China's new safeguards are still not the best time to implement them. China is faced with a growing competition, and there are sanctions against semiconductors which threaten its ability to compete in the technology sector.
Chinese A.I. raises hopes Early February, A.I. Engineer and entrepreneur Xu Liang released one of China’s first answers to ChatGPT in the form of a mobile application. ChatYuan was downloaded over 10,000 times in the first half hour, according to Mr. Xu.
There were soon media reports that there were marked differences between ChatYuan and the official party line. The responses offered a bleak assessment of the Chinese economy, and described the Russian conflict in Ukraine as "a war of aggression," which was at odds with the more pro-Russian stance of the party. The app was shut down by the authorities a few days later.
Mr. Xu stated that he would be adding measures to make a "patriotic bot". These include hiring more manual reviewers to flag problematic answers and filtering out sensitive words. He's even trained a model to detect "incorrect perspectives" which he will filter.
It is still unclear when Mr. Xu’s bot will satisfy the authorities. According to screenshots, the app was originally scheduled to resume on February 13, but it is still down as of Friday.
It read: "Service will be resumed after troubleshooting has been completed."