Altman says LLM's commoditized; China's new AI superstar Manus?; Future multipolar world
A quick wrap of the happenings in the AI world this week
Hi all,
There has been a lot of scattered information that needs to be organized and dissected together to be made sense of, so today’s post is not a deep dive or deep analysis but rather a quick news update and syndication of some current happenings in the AI world and a call for action from you - readers.
Shift in Mentality around LLMs in the U.S.
There seems to be a real shift in mentality about LLMs in the U.S. After Grok 3’ release two weeks ago, the conversation has shifted to as long as your model is “good enough,” then the real competition is now in actual adoption, and can you reach the right audience? It seems like there is some discourse between Meta’s grand ambition and its actual implementation. So far, it still hasn’t rolled out any functions that have brought on a significant amount of users for its AI agent. After all that I've written extensively about the U.S. leaning into enterprise and China leaning into consumer apps, maybeeeee we’re going to see them meeting each other in the middle.
On the BG2Pod podcast, the two avid investors in AI compared consumer AI adoption to what Search was in the internet era. There are 5 to 6 players who are clear leaders already, but eventually, it’s not just about better technology (when everyone reaches “good enough”) but more about distribution strength. Google’s self-cannabilism of its search engine seems like its smartest AI strategy yet. Still, the one who can better integrate the experience into their existing app will likely capture more users. Similarly to what I’ve been writing about, it might not be a winner-takes-all scenario; rather, a winner takes most, but then specialized vertical adoption of AI will really have its space to grow, too. And as much as the frontier model is a key competitive advantage, reach is key to capturing consumer AI. This phenomenon and advantage have been shown by Tencent’s adoption of DeepSeek in the Chinese market. See here. I’ve also written about how Tencent’s WeChat has its advantages, as X does, compared to some of its peers due to functional adjacency.
Even at the TMT conference held by Morgan Stanley this past week in San Francisco, Sam Altman of OpenAI himself has admitted that LLMs are becoming commoditized, and he’s exploring upstream and downstream - in applications and infra. When asked where value will accrue? He said at the top, it’s in applications, and at the bottom, it’ll be in chips, power, datacenters. Currently, in the middle layer, which includes models, it’s becoming hard to capture value in the long run since it's being commoditized, according to my industry source on-site at the event.
Manus - China’s new AI superstar?
Manus broke the Chinese internet yesterday. See RecodeAI Tony Peng’s quick update on this on Substack and Manu’s launch video here on X.
So far, I haven’t been able to test it out since it’s an “invitation-only web preview “ (neither has any of my mainstream media friends so far). So this information is all second-hand. The SCMP described Manus capabilities as follows: “can execute practical tasks such as creating a custom website through a step-by-step process, according to a demonstration video on its website manus.im.”
It has been hailed as a groundbreaking AI agent capable of independently executing complex tasks, marking a significant shift from traditional AI assistants that primarily provide suggestions or answers. In its launch video, it shows that it can screen CVs, provide stock analysis, and write a program to calculate budgets for your business, which seems very impressive.
Manus has been described as having a"DeepSeek moment" for AI agents, marking a transition from single-task execution (mostly prompt-to-answer chat) to intricate decision-making processes. Its ability to reason and then deliver complete task results positions it as a significant player in the AI agent space.
But while many people are raving about its capabilities and user experience, some are questioning why it feels more like a PR win for Manus than a technical breakthrough.
The launch video shows the founder, a young Chinese man, Yichao ‘Peak’ Ji (recipient of Forbes 30 under 30), speaking perfect English and walking you through the product in English. This itself has led to some netizens asking why it was not done in Chinese if the target audience is for the Chinese audience. Alternatively, many netizens in China are saying he embodies the internationalism of China’s young entrepreneurs. So we can see that, in many ways, he’s getting more attention for who he is and his background than the product itself. The whole “国运1” narrative of China’s AI is on the rise and is fueling this frenzy somewhat - which I think is going a bit out of control…
As seasoned China tech analyst Rui Ma puts it, “Don't underestimate the power of fortuitous timing,” hinting that Manus is also riding on the success of DeepSeek to garner so much public attention.
https://x.com/ruima/status/1897760289145471073
According to China Daily’s report, the invitation codes were being resold for nearly 100,000 yuan（$13,797 on the second-hand trading platform Xianyu - which is a bit wild. People are saying that the people who have had trial codes are mostly all “self-media,” meaning frankly, they could have been paid for PR, but technical people are not getting access to it. The company’s website itself is only in English as well - a bit odd. A lot of the articles are really emphasizing that Manus is a “Chinese team,” another example of China’s AI prowess. This nationalistic angle is also quite in contrast to DeepSeek’s initial coverage, as most of the conversation was about efficiency and model capabilities. I don’t have further takes on Manus, but I will be watching the development here.
A Multipolar World
These are things that are on my mind/ radar, and I hope you can contribute to them. Please leave a comment or DM me if you have any thoughts on these.
How has China’s “real economy” prioritization shaped its AI strategy? How will AI affect 实体经济 - tangible economy - manufacturing, real estate, etc -
What are some innovative ways you’ve seen vertical experts adopt AI, such as Style3D, in fashion designing
Can China and U.S. AI supremacy really not coexist? The AI arms race is being framed as a national security issue in the U.S., but it’s truly tough to “win” when it isn’t actually a binary thing. And it seems like more and more people are realizing that the “there could only be one winner” mentality is not conducive to or productive for anyone. We’re only at the infancy stage (actually, Brad Gerstner called it the “embryonic” stage) of AI development, and the world will only develop and adopt more. The doomers can keep going at it, and we need them for checks and balances, but I really don’t think somehow the world will just not continue in AI innovation.
China’s official governmental response to how it views America’s efforts in ensuring China doesn’t catch up in the AI race was rather non-aggressive. Foreign Minister Wang Yi said that high walls cannot stifle innovative thinking and decoupling or cutting off chains will ultimately isolate oneself...adding that science and technology should not be used as tools to weave an iron curtain… See the full transcript provided by . (btw
put together AI takeaways from China’s Two Sessions here).To be honest, I think the discussion should really shift to the following: Can we reach a point of international collaboration in regulating new technology or finding more efficient ways to utilize it so we don't completely destroy our earth? Wouldn’t that be much more efficient of a conversation to focus on? Can we have a multipolar world? The BG2 guys seem to think it’s quite “naive” for anyone to think that China will not achieve frontier AI simply by blocking chip exports. Mhmm, this topic deserves more time and thought. I’ll come back to it another time.
Last but not least, a little shameless self-promotional plug: check out my commentaries on China’s AI startup space and the reasons for the keen embrace of open source in the community in this SCMP article.
“国运,” national fortune, describes the overall situation and future direction of a country's development in all aspects. A good national fortune means that the country develops smoothly and is strong and influential, while a poor one indicates that the country faces numerous difficulties in development and is at a disadvantage.
Great coverage and congrats on the SCMP article!
Invitation codes now coming through with Manus and tech reviews are also rolling out. Apparently legit very impressive.