Can Xi Jinping control AI without smashing it?

Ipast For more than a month, China’s tech giants have been showing off. Baidu, Huawei, Shangtang Technology, and Alibaba are all shining in artificial intelligence (artificial intelligence) models that can power products and applications such as image generators, voice assistants, and search engines.Some have introduced artificial intelligence– Chatbot similar to Chatcommon technology, a human-like interlocutor developed in the United States dazzles users. The names of the new products are Ernie Bot (Baidu), SenseChat (Shangtang Technology) and Tongyi Qianwen (Alibaba). The latter roughly translates to the truth from a thousand questions. But in China, the Communist Party defines truth.

artificial intelligence Challenge the rulers of China. Generative sequencing processes an input of text, images, audio, or video to create new outputs of the same, and holds great promise.Chinese tech firms have been hit hard by regulatory crackdowns and the pandemic in recent years, see Generate artificial intelligence Create huge new revenue streams, similar to the opportunities created by the advent of the internet or smartphones.

However, the party believes that the generated artificial intelligence Opening up vast new avenues for the dissemination of information beyond its control. Its leaders may draw different comparisons from tech executives to the internet, which once seemed destined to help democratize China by increasing access to unfiltered news and easy communication tools. In 2000, Bill Clinton described pinning Jell-O to a wall as the party tried to take control of the Internet. Yet by deploying armies of censorship and digital roadblocks, the party has largely succeeded in creating an internet that served its own purposes and fostering an industry around it.something like this can happen artificial intelligence?

New Jelly, New Nails

The government’s concerns were made clear in the rules proposed by China’s top internet regulator on April 11. According to the State Internet Information Office (CAAC), companies should submit a security assessment to the state before using generative technologies artificial intelligence Products that serve the public. Company will be responsible for the content generated by such tools. According to regulations, the content must not subvert state power, incite secession, damage national unity, or disrupt economic and social order. And must conform to the country’s socialist core values. These restrictions may sound arcane, but similar rules apply to the internet, allowing the party to suppress speech on everything from Tibetan and Uyghur rights to democracy, feminism or gay literature.

China’s proposal comes as governments around the world grapple with how to regulate artificial intelligence. Some countries, such as the US and UK, prefer a light touch, relying on existing laws to regulate the technology. Others seem to think a new regulatory regime is needed.EU proposes a law to classify different uses artificial intelligence And apply increasingly stringent requirements according to the degree of risk. China’s approach appears more piecemeal and reactionary. Last year, for example, the party, concerned that deepfake images and videos could undermine a tightly controlled information environment, issued rules for such technologies like new ones.a clause is prohibited artificial intelligence– Resulting media has no watermarks or other clear labeling of origin.

There are other parallels to China’s approach to the Internet. Its network controls, often called the Great Firewall, may seem singular. But keeping harmful foreign content out is only part of a larger effort that has developed over time and involves many different agencies and companies. Matt Sheehan of the Carnegie Endowment for International Peace, a Washington think tank, said the first phase was laying the groundwork for an internet that the party could control.Today, the Chinese government is once again strengthening its bureaucratic muscle and adding to its regulatory toolkit, this time generatively artificial intelligence heart. Mandatory security reviews and requiring companies to register their algorithms with the state are examples of this, Mr Sheehan said.

China’s grip on the internet hasn’t stifled innovation: Just look at companies like ByteDance, the Chinese parent of the popular short-video app TikTok.But when it comes to generating artificial intelligenceIt’s hard to see how a Chinese company could create something as broad and human-like (that is, unpredictable) as Chatcommon technologywhile complying with government regulations.

this CAAC Indicates that the information generated by such tools must be truthful and accurate, and the data used to train them must be objective. The party has its own definitions of these words.But even the most advanced artificial intelligence Tools based on large language models occasionally spit out things that aren’t actually real.For a product like Chatcommon technology, which is based on hundreds of gigabytes of data extracted from all over the Internet, it is almost impossible to classify the input according to its objectivity.Strict enforcement of Chinese rules will all but prevent the development of generative technologies artificial intelligence in China.

Therefore, experts doubt that the measures will be strictly enforced. There is modest leeway in the bill. When the generated content does not meet the regulations, the government requires measures such as filtering and optimization training within three months. It sounds similar to tweaks Western companies are making to their models, for example, to stop their chatbots from spouting homophobic content or offering instructions on how to make a bomb.Local artificial intelligence Shanghai has more lenient rules, stipulating that minor violations may go unpunished at all.

the arbitrary nature of CAACThe proposed rules mean it can tighten or relax as it sees fit. Such flexibility in other countries may be seen as a good thing. But as the big internet companies can attest, the Chinese government has a habit of rewriting and selectively enforcing rules at the whim of President Xi Jinping. In recent years, companies in fields such as e-commerce, social media and video games have had to rethink their business models. In 2021, for example, state media declared video games to be psychoactive opiates, and regulators demanded that game companies stop focusing on the bottom line and instead work on reducing children’s gaming addiction.If Mr. Xi doesn’t like generating artificial intelligence He’s leaving, and he also has the ability to reset that industry.

One-way Chinese artificial intelligence Companies may be hindered by limiting the personal data that can be used to train them artificial intelligence role model. The party runs the most sophisticated surveillance state in the world, collecting vast amounts of information about its citizens. Until recently, Chinese tech companies were able to collect personal data, the lifeblood of many online businesses. But this freewheeling era appears to be coming to an end (for the private sector). Now, in theory, companies that want to use certain types of personal data must obtain consent.last year CAAC Fined the ride-sharing company Didi Chuxing equivalent to $1.2 billion for illegally collecting and mishandling user data. (The move, which was also seen as punishment for the company’s decision to list in the U.S. rather than China, was later reversed.) According to the draft rules that generated artificial intelligencethe company will be responsible for protecting the personal information of users.

plan to worry

this CAACThe proposed regulations come six years after China introduced a master plan artificial intelligence This calls for major breakthroughs by 2025 to dominate the industry by 2030. Progress towards these goals has been mixed.Chinese enterprises in the following fields artificial intelligence-Assisted image recognition and autonomous driving, which are doing well given the government’s far less focus on social stability. They benefit from substantial public funding—indeed, some provide the tools to support state repression. But China still lags behind the United States in investment and innovation.Potential foreign backers turned away by U.S. sanctions on some Chinese firms artificial intelligence giant. To make matters worse, the United States has restricted exports to China of the type of cutting-edge semiconductors that power energy. artificial intelligencea move that could hamper the industry.

On the regulatory front, China may have more success.Under the master plan, the country will write a code of ethics for the world artificial intelligence by 2030.This is an extension, but it is useful for generating rules artificial intelligence More detailed and broader than has been suggested elsewhere, thus influencing the debate on how to deal with new technologies. China’s ability to quickly enact new regulations means other countries will learn from them. One risk is that it becomes overpowering and hinders innovation. But Jeffrey Ding of George Washington University points out the other side of the debate. Noting the ingenuity of Chinese Internet companies, he said that sometimes restrictions on creativity spur more innovation.

Nonetheless, China may be involved in artificial intelligence Morality should terrify Western governments. They may have some of the same concerns as China, including misinformation and data protection, but for different reasons. Likewise, China’s experience with the Internet is beneficial. It firmly rejects the notion of the Internet as a free and open place. When governments meet to discuss online regulation, China has consistently sided with Russia and other free-speech abusers. Mr. Clinton naively believes that the Communist Party cannot bring the Internet to its knees. It would be naive for Western leaders today to think it cannot do the same with China. artificial intelligence.