The Chinese startup Butterfly Effect, which has developed the general-purpose artificial intelligence agent dubbed Manus, appears to become a significant change factor in the market of large language models in the Asian country, as this virtual product focuses on apps beyond chatbots similar to ChatGPT from OpenAI.
The mentioned startup backed by Tencent Holdings found itself in the area of increased attention from the AI community following its invitation-only online preview last week for Manus. This virtual product can perform various practical tasks, such as creating a custom website.
The startup claims that Manus is superior to other chatbots in terms of performance. This virtual product, among other things, can quickly create a trip itinerary to Japan, conduct an in-depth analysis of Tesla shares, and offer real estate tips in New York, focusing on the needs of the user’s family.
Butterfly Effect founder and chief executive officer Red Xiao Hong stated that the artificial intelligence agent developed by the startup looks more like a human compared to other chatbots. According to him, Manus not only thinks and answers questions, but also interacts with its environment, collects feedback, and uses these information materials as a new prompt.
During a conversation with media representatives, Red Xiao Hong stated that he had not considered the development of large language models from day one due to the desire to focus on apps, which are a segment of the artificial intelligence industry. In his opinion, the mentioned segment has still not matured. Large language models are the technology underpinning generative artificial intelligence services such as ChatGPT.
Butterfly Effect co-founder and chief scientist Peak Ji Yichao stated in a post on X that Manus was built on existing large language models, including Claude from Anthropic and fine-tuned versions of Qwen from Alibaba Group Holding.
Red Xiao Hong, who often shares his thoughts on the Chinese social media platform Jike, stated in his latest post that he’s busy expanding computing capacity and fixing bugs.
The commitment to apps demonstrated by Butterfly Effect shows another avenue for innovation by developers from an Asian country in the artificial intelligence industry after DeepSeek’s breakthrough this year in building AI models at a fraction of the cost and computing power.
Red Xiao Hong’s opinion partially echoes that of Robin Li Yanhong, the co-founder, chairman, and chief executive officer of Baidu, who called on Chinese technology leaders to pay more attention to elaborating real-world apps powered by machine intelligence, according to a panel discussion at the World Artificial Intelligence Conference in Shanghai in July. According to him, in 2023, intense competition among over 100 large language models emerged in the Asian country, which led to a significant waste of resources, especially computing power.
After the debut of ChatGPT in November 2022, which was actually the initial stage of the so-called artificial intelligence boom, both Big Tech companies and startups mainly focused their efforts on large language models, the application programming interface of these models for developers, and chatbots for individual and enterprise users.
The race to train large language models led to significant investments in computing power and a price war. The companies involved in the relevant process pushed to monetize their products.
Following the success of DeepSeek, Chinese technology firms are looking to raise new funds and push their open-source large language models to the market.
Open source provides public access to a software program’s source code. In this case, third-party developers can modify or share the design of the mentioned models. They can also fix broken links or scale-up capabilities.
Chinese media reported that ByteDance, which ownership structure includes TikTok and Douyin, tried to acquire the developer of Manus at the beginning of last year as part of a deal worth $30 million, but this offer was turned down.
ByteDance and Butterfly Effect did not respond to a request for comment regarding the mentioned information.
The Manus developer claims that this virtual product outperforms OpenAI’s Deep Research based on the GAIA benchmark, a third-party measure of general AI assistants. The mentioned AI agent preview sparked significant interest across the artificial intelligence community. At the same time, the limited availability of the app has already caused some doubts about the capabilities of Manus.
Manus product partner Zhang Tao said in a Jike post last week the current invite-only mechanism is due to limited server capacity at the present stage.
Red Xiao Hong founded Butterfly Effect two months before the launch of ChatGPT. The startup’s first product was an artificial intelligence assistant called Monica.
It’s worth noting that during the period of active development of machine intelligence, the issue of cybersecurity has become more relevant. In the context of the corresponding circumstance, it is important to underline that fraudsters also have access to artificial intelligence tools. Against the background of the practice of using the mentioned tools, their activities have become more sophisticated. To counteract the corresponding threat in a virtual environment, personal awareness of users is important. For example, an Internet search query such as how to know if my camera is hacked will allow anyone to get information about signs of unauthorized access to the device. Digital literacy is an effective tool for countering cybercrime. At the same time, this knowledge should be updated periodically because scammers strive to use the most advanced technological solutions in their activities.
