AI race will not be played without China. Less than a month after the excitement aroused by the Chinese conversational robot Deepseek, the Middle Empire launched a new competitor to Chatgpt on March 6. His name: Manus.
Developed by the start-up Butterfly Effect, this artificial intelligence agent wants to be as ingenious as humans. But if manus is, for the moment, only available by invitation, its existence already arouses curiosity. And above all … worries.
An AI capable of “buying a property”
“It’s not just another chatbot. He is a truly autonomous agent,” said his creator Yichao “Peak” Ji in a promotional video. “Where the others AI just generate ideas, Manus brings results. We see it as the next paradigm of collaboration between men and machines, and potentially an overview of the general AI (as intelligent as humans, editor’s note),” he added.
On its website, Butterfly Effect praises the tasks that manus is supposed to be able to accomplish. Among them: analyze Tesla’s action on the stock market, buy a property in New York, edit a podcast, or even organize a trip to Japan. But this ideal is still far from being reached. In an article published on Sunday, Kyle Wiggers, journalist at Techcrunch, notes that the IA agent failed to order a sandwich and find him a plane ticket towards the land of the rising sun.
https://www.youtube.com/watch?v=undefined
At the heart of concerns, data confidentiality
Beyond his imperfect capabilities, the sudden launch of manus concerns. The Silicon Valley, the first, worried about it. In question: the rapid catch -up of China in the AI sector, despite the American restrictions on the export of advanced flea.
Each day, the improvement of the AI agents is more feared from the drifts linked to the lack of regulation of the new technology. Manus does not escape this rule. “The models are making many factual errors. If they are entrusted with tasks with high stakes, such as the purchase and sale of actions, these imperfections could lead to chaos,” alerts Mel Morris, director of Corpora.ai, a search engine at generative AI. “His ability to access remote servers raises the same concerns about data confidentiality,” he continues.
The launch of manus is all the more worried as the controversy aroused by its Chinese predecessor Deepseek remains in the spirits. At the end of January, Italy had blocked the use of the R1 conversational robot on its territory, invoking a lack of transparency on the processing of personal data. Taiwan, Australia and South Korea had walked in its steps, fearing for national security or potential leaks of sensitive information.