When Open AI unveiled ChatGptlast November, two immediately clicked “red codes”. The former rocked Google’s offices in Mountain View, California. The second alarm, on the other hand, caused quite a stir in Beijing, that is to say on the other side of the world.
In the headquarters of Baiduthe Chinese multinational specialized – among other things – in artificial intelligence, the Cto Wang Haifeng has given a peremptory order: complete the tests of generative AI by March 2023. And have a product similar to ChatGpt integrated as soon as possible in the search engine bearing the company name.
Baidu is the most used search engine in China, chosen by 80% of surfers. Almost all of the profits of the Chinese multinational depend on its performance, which for some time has been heralding a model of large-scale language, chiamato Wenxinbased – according to company communications – on approx 260 billion parameters. A number that far exceeds i 175 billion parameters that make up Gpt 3the deep learning algorithm on which ChatGpt was trained.
The Chinese answer to ChatGpt – based on Wenxin – his name will be Ernie and apparently it will debut in March 2023as requested by Wang Haifeng.
In Beijing they have been working on it since 2019. Even earlier, in 2017, Baidu opened – in collaboration with the Chinese government – an important research laboratory dedicated to the development of artificial intelligence. It includes scientists and academics who once worked at the development of military robots.
Just a few years ago, China seemed close to overtaking the United States in the AI race. In September 2021, Wipo, the United Nations agency specializing in the protection of intellectual property, communicated that China led the world in terms of AI-related patents.
At the time the Chinese advance started to scare America. In June 2021, the Senate approved a bill that aimed to increase the country’s resources to compete with the technology arriving from Asia.
Yet for more than one reason, today, Baidu is also chasing ChatGpt and a small startup from San Francisco – Open AI – that has bewitched the world.
The first example of generative AI created by Baidu, called Plato and launched in 2021, it did not impress. She was unable to write correctly, for example, even the birth dates of Chinese celebrities.
The difficulties that a new technology can encounter – we have seen what happened to Meta’s Galactica, Google’s Bard and Microsoft’s new Bing – join in China complications caused by government interferencewhich for several years has begun to control the country’s big tech companies in an increasingly influential way.
Artificial intelligence
The new Bing (which uses ChatGpt) already makes mistakes. But it shouldn’t surprise us
by Pier Luigi Pisa
In China today there is a fairly eloquent saying: “We have to teach machines not only what to say, but also what they absolutely must not say.”
Despite this, Chinese researchers are working to offer their compatriots an alternative to ChatGpt, which – we remind you – is not accessible (officially) in the great Asian country. Indeed, the Government has communicated to the Chinese tech companies of not offer services based on the generative AI developed by the US Open AI to the public.
The fear is that a tool similar to ChatGpt, capable of generating unpredictable responses, could produce content capable of escaping censorship.
Fake news
ChatGPT could take the spread of disinformation to levels never seen before
edited by Jack Brewster, Lorenzo Arvanitis and McKenzie Sadeghi
Maybe that’s why it was closed almost immediately Mossa generative artificial intelligence similar to ChatGpt unveiled on February 20th.
Moss was developed by Natural Language Processing Lab of Fudan University in Shanghai, part of the “C9 League” which brings together the best universities in the country in terms of education and above all innovation.
Xipeng QiuFudan computer science professor, said Moss can be considered a simpler version of ChatGptas the number of parameters Moss was trained on – 11 million – is significantly fewer than those used by Open AI.
A few hours after the launch of Moss, the servers on which the chatbot was based collapsed. The researchers didn’t expect such a high peak of requests. This, at least, was the official reason that led to the cancellation – for an indefinite time – of the beta version of the AI.
On the page hosting the Moss chatbot, it currently says this: “To ensure a better user experience, MOSS needs to be updated and the service is now down”.
Meanwhile, the Chinese government makes itself heard, reassuring.
According to Ministry of Science and Technology of Chinathe country has a technology similar to ChatGpt that will be able to integrate artificial intelligence into society and economy.
“This technology will be applied in many industries and fields,” he said Chen Jiachangwho heads the ministry’s tech department, at a recent news conference.