Recently, an AI tool called Manus suddenly caught fire all over the Internet, with some people calling it the "iPhone of the AI world" and others saying it is just another "PPT-style AI". As an ordinary person concerned about the development of AI, I spent three days researching relevant information, trying to unveil this mysterious tool.
What is Manus?
Simply put, Manus is aThe "007" of AI.-- It not only answers questions, but also performs a variety of complex tasks like a secret agent. For example:
- You say "make a PPT" and it generates the full file directly
- You say "write a game" and it develops and deploys a web version of Doom
- You say "analyze stocks" and it automatically organizes the data and generates visual charts
This is in stark contrast to ChatGPT, which is all chat. Just like when you ask your secretary to write a report, ChatGPT will tell you how to write it, whereas Manus will just write the report and give it to you.
Second, does it really outperform OpenAI?
Here's a recent bombshell: Manus outperformed OpenAI's DeepSeek in the prestigious GAIA test, which is like the AI world's "college entrance exam," looking at the ability to complete tasks. But there are two key points to note:
- Theoretical results ≠ practical ability: Just like the schoolteacher who scores high on tests but has poor hands-on skills, DeepSeek has been criticized for scoring high but being slow.
- Limitations of the internal testing phase: Currently only open for testing to a limited number of people, demo cases may be carefully selected
What's more, the real AI gurus (such as the OpenAI team) have not publicly commented on it so far, which reminds me of Sora, which was a big hit last year - blown out of the water at the time, but still has a long way to go to commercialization.
III. The team's "identity paradox"
The most interesting thing about Manus is its "identity mystery":
- Claims to be a Chinese team, but CEO launches product in English
- The design of the official website is completely European and American, even the Chinese text is very little!
- Avoid mainstream media when releasing and choose niche channels to hype it up
There are two possible reasons behind this maneuver:
- Financing Strategies: wants to attract Silicon Valley capital (cf. OpenAI's path to success)
- Regulatory Considerations: Avoiding pressure from domestic public opinion
However, this "eat Chinese traffic, earn international capital" approach, it is easy to fall into the "inside and outside" of the embarrassing situation.
Fourth, what is the real experience really like?
Although I haven't gotten my hands on the beta yet, I can summarize the issues by analyzing the feedback that's already out there:
- technical shortcoming: Sites with CAPTCHA such as Zhihu are inaccessible and complex tasks get stuck easily
- utilization threshold: need to manually adjust the parameters, not friendly to ordinary users
- Stability in doubt: Processing long tasks (e.g., organizing industry reports) often fails
These issues are common in the early stages of AI, but must be addressed soon if it is to become a mainstream tool.
V. Will it change the world?
In the long run, Manus is on the right track:
- The Revolution of the AI Agent: Evolving AI from 'tool' to 'assistant' could accelerate AGI's arrival
- Efficiency gainsThe "AI worker" industry could emerge in the future, as ordinary people can perform specialized tasks with ease.
But in the short term, we need to remain rational:
- Don't follow blindly, wait for large-scale test results
- Beware of 'AI Anxiety', Human Creativity Remains Irreplaceable
In conclusion, I'd like to say this: if Manus can really deliver what's in the demo, it's truly revolutionary; but if it's just another "PPT project", we don't have to be too disappointed -- after all, in the field of AI, every year a few After all, in the field of AI, there are a few "disruptors" every year, and what really changes the world is often the technology that quietly persists.
