Bill Clinton once quipped that Beijing's attempts to control the Internet would be «like trying to nail Jelly to a wall.» . The former US president optimistically predicted that the freedom of the world wide web will force China to open up as citizens have greater access to information.
Clinton underestimated the Chinese Communist Party. The country's great firewall has cut off access to Western websites, and cyberspace regulators strictly control what is said on its domestic counterparts. Instead of weakening the power of the government, Beijing's control of the Internet has strengthened it.
But if Internet censorship was like nailing jelly to a wall, trying to manipulate artificial intelligence (AI) could be like one-handed back.
The emergence of so-called «large language models» such as ChatGPT, which collect massive amounts of data and confidently produce human-sounding answers, represents a new challenge to Beijing's ruthless Internet censorship.
Alibaba CEO Daniel Zhang presents his company's response to the ChatGPT craze. Photo: Alibaba
The Cyberspace Authority of China (CAC), the country's powerful internet regulator, said on Tuesday that generative AI such as chatbots and imaging systems will be subject to stringent security checks that ensure any content they create is loyal. government.
“Content created using generative AI must embody core socialist values and must not contain material that undermines state power, advocates the overthrow of the socialist system, incites the division of the country, or undermines national unity,” the CAC said in a statement.
Companies such as Alibaba, SenseTime and Baidu will stress-test their chatbots with regulators in an effort to provoke them to criticize Xi Jinping and spark a revolution.
The announcement comes on the same day that internet giant Alibaba, which has been under scrutiny by the Chinese government for years, said it would release a chatbot called Tongyi Qianwen, which means «the truth from a thousand questions.»
Western chatbots such as OpenAI's ChatGPT, Microsoft's Bing and Google's Bard are already blocked at the website level in China, and the Chinese government has also banned companies from using OpenAI technology in their services. The state-run China Daily warned that ChatGPT could «give a helping hand to the US government in its dissemination of disinformation and manipulation of global narratives for its own geopolitical interests.»
Is artificial intelligence developing too fast? Poll
China has become an AI superpower. In 2017, the country laid out plans to become the world's premier artificial intelligence hub by 2030. The push was reportedly made after a bot developed by DeepMind, Google's UK subsidiary, beat the world champion in the hellishly difficult Chinese board game Go. .
By last year, Chinese researchers had published three times as many papers on artificial intelligence as American researchers, according to Ohio State University researchers. a key element of a country's military strategy.
Large-language models such as ChatGPT have been described as an Internet-scale economic catalyst that allows white-collar workers to outsource much of their work to machines.
However, China's strict control over them may hamper development. “Here in the West, governments are pretty quiet about all the inaccuracies of misinformation, hallucinations and things like ChatGPT,” says Stephanie Hare, technology researcher. “[But] Alibaba can't work the same as OpenAI. Western companies that enjoy more freedom and less political interference than their Chinese counterparts are likely to pull ahead.”
Bots like ChatGPT are trained to work with huge amounts of data and act like advanced versions of Google's autocomplete, producing snippets of text that are incredibly human-like. However, software is notoriously unpredictable, capable of spouting lies, making up facts, or, with the right prompts, ignoring their software to become aggressive. Last week, the Mayor of Australia threatened to sue ChatGPT for defamation after he said he spent time in prison.
Fraudulent Chinese AI has the potential to have serious repercussions for its developer, meaning that chatbots are built with strict restrictions. Try asking Ernie, a chatbot developed by search engine Baidu, about Xi Jinping, and he will reply that the question «cannot pass security checks.» Click it further and it will ask you to change the theme to something lighter.
Baidu CEO Robin Li speaks at the presentation AI chatbot Ernie Bot at an event in Beijing. Photo: Michael Zhang/AFP
However, this is only a temporary solution. China used the internet as a way to bolster communist rule, and the CAC is pushing for chatbots to reflect socialist values, not just ignore questions about them.
Richard Windsor, an independent technology analyst, says these requirements, plus US trade bans on China's access to the most advanced microchips, mean that Chinese chatbots will be technologically inferior if they are only trained on approved data. The easiest way to avoid giving the wrong answer about Tiananmen Square, for example, is to make sure «ChatCCP» doesn't know anything about it.
«The Chinese will have to use much smaller datasets that have been very carefully curated. OpenAI [and others] don't have these restrictions and so they can just download the entire data set from the internet into the system and see what comes out on the other end,” he says.
However, Windsor says it could give developers an easier understanding of chatbots, making them more reliable. «The CAC regulation will help get rid of the craziness … but not the usefulness that will make these models economically viable,» he adds.
Arthur Holland Michel, senior fellow at the Carnegie Ethics Council at International Affairs, says, that regulation of AI systems in the US and Europe is inevitable, meaning that earlier Chinese regulations could give developers a head start in working within strict limits.
“In the West, we are in the pre-regulation stage. I don't think anyone has fully grasped the possibility that after you include this material in [legislation], some of the same questions will arise,” he says.
If Chinese Chat If the bot takes off in the West, it could raise fears that will make the TikTok panic seem irrelevant. Beijing's restrictions on independent thinking coming from human or machine means this is unlikely to happen.
Свежие комментарии