Michael James

5.9K posts

Michael James banner
Michael James

Michael James

@Islaladrones

Building AI applications for schools. Director The Charles Prosser Foundation. Asia based. Stanford. Occasional thoughts on the AI industry.

Hong Kong/Taipei Katılım Nisan 2022
847 Takip Edilen212 Takipçiler
EdSurge
EdSurge@EdSurge·
Neuroscientist Jared Cooney Horvath argues in his new book "The Digital Delusion" that digital devices and screen time interfere with children's cognitive development — at the expense of play. bit.ly/4dL33IQ
English
1
1
4
370
Michael Pettis
Michael Pettis@michaelxpettis·
15/15 imbalances, there will only be two ways this can be aggressed. Either we move to a world in which no major economy controls its trade and capital account, or we move to a world in which every major economy exerts increasing control over its trade and capital account.
English
6
0
39
5.5K
Michael Pettis
Michael Pettis@michaelxpettis·
1/15 IMF: "If coordination proves difficult, the best course of action for each country is clear: start addressing domestic imbalances now, regardless of what others do." This is one of several discordant lines in an otherwise interesting paper. imf.org/en/blogs/artic…
English
5
31
143
51.6K
John Ross
John Ross@JohnRoss43·
Yu Yongding: "There Is No 'Consumption-Driven' Growth Model, and China’s Infrastructure Investment Is Far From Saturated." Outstanding article by one of China's leading economists cuts right through the economic confusion and factual and theoretical non sequiturs in much of the Western media. Compulsory reading for anyone who wants to seriously understand China's economy. eastisread.com/p/yu-yongding-…
English
17
138
454
50.9K
Lars Christensen
Lars Christensen@MaMoMVPY·
I am increasingly coming to this conclusion - there NO SIGNS the problems with hallucinations are getting solved. In fact if anything it is now spreading to coding and the use of agents where it is hidden and could led to serious problems down the road. Therefore we can't really scale LLMs. LLMs are very useful tools, but you need to know about the limitations. Most of the investments in LLMs today assumes away these limitations.
Merryn Somerset Webb@MerrynSW

What if the whole LLM thing is a false start? If the flaws are inherent systemic problems - if the compounding of hallucinations/errors can't be sorted out? If the capex build out is one of the biggest misallocations of capital ever? Then what? bloomberg.com/news/newslette…

English
132
108
923
122.4K
Michael James retweetledi
Mike
Mike@Doranimated·
This posting is brilliant
James E. Thorne@DrJStrategy

Food for thought. Trump, Hormuz and the End of the Free Ride For half a century, Western strategists have known that the Strait of Hormuz is the acute point where energy, sea power and political will intersect. That knowledge is not in dispute. What is new in this war with Iran is that the United States, under Donald Trump, has chosen not to rush to “solve” the problem. In Hegelian terms, he is refusing an easy synthesis in order to force the underlying contradiction to the surface. The old thesis was simple: the US guarantees open sea lanes in the Gulf, and everyone else structures their economies and politics around that free insurance. Europe and the UK embraced ambitious green policies, ran down hard‑power capabilities and lectured Washington on multilateral virtue, secure in the assumption that American carriers would always appear off Hormuz. The political class behaved as if the American security guarantee were a law of nature, not a contingent choice. Their conduct today is closer to Chamberlain than Churchill: temporising, issuing statements, hoping the storm will pass without a fundamental reordering of their responsibilities. Trump’s antithesis is to withhold the automatic guarantee at the moment of maximum stress. Militarily, the US can break Iran’s residual ability to contest the Strait; that is not the binding constraint. The point is to delay that act. By allowing a closure or semi‑closure to bite, Trump ensures that the immediate pain is concentrated in exactly the jurisdictions that have most conspicuously free‑ridden on US power: the EU and the UK. Their industries, consumers and energy‑transition assumptions are exposed. In that context, his reported blunt message to European and British leaders, you need the oil out of the Strait more than we do; why don’t you go and take it? Is not a throwaway line. It is the verbalisation of the antithesis. It openly reverses the traditional presumption that America will carry the burden while its allies emote from the sidelines. In this dialectic, the prize is not simply the reopening of a chokepoint. The prize is a reordered system in which the United States effectively arbitrages and controls the global flow of oil. A world in which US‑aligned production in the Americas plus a discretionary capability to secure,or not secure, Hormuz places Washington at the centre of the hydrocarbon chessboard. For that strategic end, a rapid restoration of the old status quo would be counterproductive. A quick, surgical “fix” of Hormuz would short‑circuit the dialectic. If Trump rapidly crushed Iran’s remaining coastal capabilities, swept the mines and escorted tankers back through the Strait, Europe and the UK would heave a sigh of relief and return to business as usual: underfunded militaries, maximalist green posturing and performative disdain for US power, all underwritten by that same power. The contradiction between their dependence and their posture would remain latent. By declining to supply the synthesis on demand, and by explicitly telling London and Brussels to “go and take it” themselves, Trump forces a reckoning. European and British leaders must confront the fact that their energy systems, their industrial bases and their geopolitical sermons all rest on an American hard‑power foundation they neither finance nor politically respect. The longer the contradiction is allowed to unfold, the stronger the eventual synthesis can be: a new order in which access to secure flows, Hormuz, Venezuela and beyond, is explicitly conditional on real contributions, not assumed as a right. In that sense, the delay in “taking” the Strait, and the challenge issued to US allies to do it themselves, is not indecision. It is the negative moment Hegel insisted was necessary for history to move. Only by withholding the old guarantee, and by saying so out loud to those who depended on it, can Trump hope to end the free ride.

English
28
59
282
90.6K
Michael James retweetledi
#FreeJimmyLai
#FreeJimmyLai@SupportJimmyLai·
At the National Prayer Breakfast, Claire Lai shared her father’s story. In 1997, the year Hong Kong was handed to China, hearts across the city were filled with fear and doubt. #JimmyLai turned to Christ. His doubts were overcome by God’s mercy and grace and transformed into trust. This Easter, Claire says there is nothing he wants more than to be a faithful servant to our Lord. He remains alone in a cell, sentenced to 20 years, for refusing to stop believing in freedom. #FreeJimmyLai
English
6
51
197
5.9K
Anthony Bradley
Anthony Bradley@drantbradley·
I recently heard someone equate college students using AI as equivalent to using calculators. I laughed. This is not what we’re seeing in the humanities and social sciences. Students are using AI to avoid learning and deep-think writing assignments. It’s not a “tool.”
Anthony Bradley tweet media
English
67
196
891
43.4K
Kyle Chan
Kyle Chan@kyleichan·
“China’s relatively high level of investment is not a sign of distortion or imbalance, but rather one of the main reasons the country has sustained rapid long-term growth.” Provocative piece by @ywang2005b, former World Bank senior economist: bu.edu/gdp/2026/03/27…
English
7
49
144
22.8K
Anand Sanwal
Anand Sanwal@asanwal·
Wharton researchers gave nearly 1,000 high school math students access to ChatGPT during practice problems Result: chatGPT is the perfect trap. Look at the red bars. Students with ChatGPT crushed their practice sessions. The basic ChatGPT group solved more problems and those on the "tutor" version did even more. Now look at the gray bars. That's the exam. No AI allowed. The ChatGPT group scored 17% worse than kids who practiced with zero technology. And the fancy tutor version? No better than working alone. The researchers called AI a "crutch." When they analyzed what students actually typed into ChatGPT, most of them just wrote - “What’s the answer?” The kicker: students who used ChatGPT believed it hadn't hurt their learning. They were confidently wrong. This is the AI trap in education. Outsourcing your thinking. Of course, lots of half-baked AI literacy curricula being rolled out in schools now Let’s of course ignore that basic literacy (the ability to read) is possible for <50% of 8th graders Source: Bastani et al. (2025), "Generative AI Can Harm Learning," PNAS
Anand Sanwal tweet media
English
217
1.2K
4.1K
752.5K
Gerard DiPippo
Gerard DiPippo@gdp1985·
Comparing China's GDP to the world (or the US) in nominal dollars invites criticisms about which exchange rates to use. In 2026, the RMB has strengthened against the USD so far, so China's share of global GDP will likely increase. The more remarkable story is that China's nominal GDP growth is lower than that of the EU, Japan, and the US, even when measured in local currency terms.
Gerard DiPippo tweet media
Yanzhong Huang@YanzhongHuang

China's churning out more goods than ever—EVs, solar, ships, exports booming—yet its GDP share of the global economy in nominal dollars keeps shrinking. Peaked ~18.5% in 2021 → now ~16.6%. If the trend holds, China won't catch up to (let alone surpass) the US economy by 2035. Big dent in the "rising superpower" narrative. wsj.com/world/china/be…

English
8
12
53
9.9K
Wolfcop
Wolfcop@Wolfcop9·
@csubagio @Islaladrones @natashajaques There's the story about New Coke in the 80s. When they did taste tests, people preferred it to the original formula. But when it went to market, sales dropped. It turned out people liked drinking original Coke consistently. New Coke wasn't something they could drink all day.
English
2
0
1
17
Natasha Jaques
Natasha Jaques@natashajaques·
The paper I’ve been most obsessed with lately is finally out: nbcnews.com/tech/tech-news…! Check out this beautiful plot: it shows how much LLMs distort human writing when making edits, compared to how humans would revise the same content. We take a dataset of human-written essays from 2021, before the release of ChatGPT. We compare how people revise draft v1 -> v2 given expert feedback, with how an LLM revises the same v1 given the same feedback. This enables a counterfactual comparison: how much does the LLM alter the essay compared to what the human was originally intending to write? We find LLMs consistently induce massive distortions, even changing the actual meaning and conclusions argued for.
Natasha Jaques tweet media
English
47
390
1.5K
251K
Cham
Cham@Marxistcham·
@Islaladrones @GlennLuk Dude read some report-card analysis of the 14th FYP and what it has been able to achieve. Don't be lazy and repeat common tropes like 'overcapacity' or 'lack of consumption'. Even the most anti-China press would concede 14th FYP as 'limited success".
English
1
0
1
36
Glenn
Glenn@GlennLuk·
China's focus on technology and industrial development is not new, as articulated below in Kyle's survey of historical 5YPs. But why it is increasingly impactful has less to do with intent but the sheer quantity of relevant (i.e. high levels of formal education) human capital that it now has to allocate across an expanding set of disciplines.
Kyle Chan@kyleichan

🚨 My NEW High Capacity piece 🚨 CHINA'S TECHNOLOGY LONG GAME What reading through China's Five-Year Plans reveals about its evolving tech strategy--and its plans for the industries of the future.

English
6
40
167
14.9K
Greg R. Lawson
Greg R. Lawson@ConservaWonk·
McCarthy is like stone. Permanent in an age of impermanence. Though I have only read Blood Meridian because of its reputation - I think, along with Conrad's Heart of Darkness, are two of the greatest meditations on the evil that man is truly capable of in a world devoid of faith. For in that world, Kurtz & Judge Holden would not be villains but realists. This is why I always urge reading #Dostoevsky, especially Crime & Punishment, The Brothers Karamazov, & The Demons, alongside #Nietzsche & things like McCarthy & Conrad. A world without faith is not a world of decency & love but of brutality, wanton cruelty & the Will to Power. #Literature #Philosophy #Faith
Bandy@bandy1803

It makes no difference what men think of Cormac McCarthy's books. They endure. As well ask men what they think of stone.

English
1
1
11
541
Ryan Hass
Ryan Hass@ryanl_hass·
Sharing my contribution to @ForeignAffairs. I argue that US and China have entered a period of fragile stability. What each country does to strengthen itself during this window will have more impact on direction of relations than what Trump and Xi discuss together. Link below.
Ryan Hass tweet media
English
9
15
54
8.4K
Tom Loveless
Tom Loveless@tomloveless99·
SFUSD has it backwards. It makes sense to offer struggling students a "double dose" of math--to help them catch up. But to double dose advanced students, with one of the math classes below their level of math, makes no sense.
English
5
4
44
1.6K
Glenn
Glenn@GlennLuk·
Now that energy prices are on the rise, I just can't wait for all the China deflation doomers to suddenly switch to being inflation doomers and change the filter on their GDP watching glasses from nominal to real.
Gavekal@Gavekal

As the Iran war continues, higher commodity prices will mechanically lift China's headline inflation. This will likely push the PPI out of deflation and could turn the GDP deflator positive, fulfilling a government goal to reverse price declines.

English
5
14
168
9.2K
Michael James
Michael James@Islaladrones·
@slow_developer Uh. No. Alibaba, XiaoMi and others all moving away from open weights. The chapter of the story is ending.
English
0
0
0
9
Haider.
Haider.@slow_developer·
wow, i didn't expect this from elon but he basically admitted two big things: "china is leading the AI race globally, and google is leading it in the west" yes, in the end, open-source AI wins. open-source models will only be 3–4 months behind the top labs -- but they have a bigger base, grow faster, and hold the stronger long-term edge
Haider. tweet media
English
81
25
337
39.3K