The Japan Times - What to know about Manus, China's latest AI assistant

Manus, an AI agent generally considered more advanced than a chatbot, can do everything from analysing the stock market to creating a personalised travel handbook for a trip with simple instructions from users, its website says.

Here's what you need to know about Manus:

- Rapid rise -

Manus was released last week by Chinese startup Butterfly Effect.

In a now viral introductory video posted online, co-founder Yichao "Peak" Ji hailed it as "the next paradigm of human-machine collaboration, and potentially a glimpse into AGI," referencing artificial general intelligence that aims to think the way humans do.

Currently accessible only by invitation, Manus has quickly gained traction, with its official Discord server growing to more than 170,000 members.

Its name comes from the Latin motto "Mens et Manus", meaning "mind and hand" -- a nod to its integration of knowledge and practical application.

- Exclusive access -

Manus is a "system that can carry out tasks autonomously on behalf of users", said Manoj Harjani, a research fellow at Singapore's S. Rajaratnam School of International Studies (RSIS).

"DeepSeek and ChatGPT differ from Manus in that they provide responses to prompts from users through a chat-style interface, whereas Manus is able to execute tasks such as booking tickets and sorting through resumes," he explained.

And while DeepSeek was quietly developed behind the scenes before rising to prominence, Manus is taking a different route -- limiting access to an invite-only beta and targeting enterprise-level clients, one analyst said.

"While this exclusivity can generate buzz, it may also impede widespread adoption," said Marina Zhang, an associate professor at the University of Technology Sydney's Australia-China Relations Institute.

She warned Manus may not cultivate the same broad appeal as DeepSeek, which built a large community through open-source engagement, if it continues to retain a closed ecosystem.

When AFP tested Manus with an invitation code given by the company, the AI assistant took significantly longer than DeepSeek to generate responses.

But it was able to complete more difficult tasks compared to its Chinese rival or ChatGPT, like creating custom websites.

- Touchy topics -

And while DeepSeek said it was "programmed" to provide answers that toe the government line on topics considered politically sensitive in China, Manus was able to give accurate, non-censored responses.

On the topic of the bloody 1989 crackdown on pro-democracy protesters in and around Tiananmen Square in Beijing, Manus responded by stating that "the Chinese government carried out a violent crackdown on pro-democracy protesters in Tiananmen Square in Beijing", followed by six more paragraphs detailing the incident.

It also gave a comprehensive answer when asked about allegations of human rights abuses by Beijing in Xinjiang.

Asked about its uncensored replies, the programme said it does not "intentionally censor factual information".

"When discussing sensitive topics like historical events or human rights issues, I aim to provide balanced, objective information that helps users understand complex subjects," it added.

This could likely be because Manus "is focused on tasks and the team didn't build content control as thoroughly as chatbots like DeepSeek and ChatGPT", said Li Jianggan, founder of Momentum Works, a Singapore-based consultancy focused on tech companies.

- Next DeepSeek? -

Whether Manus can achieve mainstream success like DeepSeek will depend on its ability to scale to meet demand, RSIS's Harjani told AFP.

But Manus is unlikely to be the next DeepSeek as they are different types of AI applications, he added.

Its ability to grow would rely on "adequate computing power and effective handling of potential challenges such as technical stability and ethical or regulatory considerations", Zhang told AFP.

"If Manus can successfully address these issues... it could indeed be a major player in the enterprise automation space," she said.

But it remains "far from perfect" and is likely not open to the wider public yet as the team may still want to make improvements to the programme, Li said.

"It is dealing with real world problems and executing real world tasks, which are diverse and complex," he told AFP.

Its success remains a "big question mark".

M.Saito--JT