@Andre_Kostitsin @Scholars_Stage Which ones? The schools I know about in shanghai don’t have that.
English
dots
94 posts
@d0t2_ @Scholars_Stage You have to write entrance exams to get in
English
As someone who used to teach at one of these schools this is absolutely ridiculous. The top 10 or so international schools in Beijing and Shanghai are so selective that it does not *matter* what you teach the kids. The top of the classes would be world class talents if you scheduled an hour of crayon drawing everyday.
LinaHua@Linahuaa
DeepSeek is reportedly auto-rejecting Chinese applicants who went to America for undergraduate. Having an undergrad from America usually means you went through the international school pipeline, which usually means that you had much less rigorous math training. It also means you're culturally White, and probably a bad fit for them. I have lots of students from Shanghai American School (#1 international school in the city)- and they have tons of bullshit filler classes like choir, abstract art, reading books about disabled/morbidly obese kids and writing essays about why fatshaming is bad. As the result- they're far behind in math and science. And they're somehow less competent at music and arts too because they skip the foundation and go straight to vibe-painting and virtue-yapping. American education is of very low quality imo... this isn't a China glazing post btw- I would take the average German grad from 20 years ago over today's American grads too. I would also take the average German (from 20 years ago) over the average Singaporean and Chinese.
English
@AnalyticaCamil1 All this would at best put him in upper middle class. Do you know how dime-a-dozen those are? And it was a *LOAN*, made when he was *30*, and already hyper-successful at DE Shaw, where he could’ve made $10-100Ms. And tf you mean no business outlays? You think Amazon is a SaaS?
English
Leaving out that Jeff Bezos’ mom immediately married a Cuban-American engineer who wound up working at peak-Exxon, and that the two had the capital to give him a quarter million dollar loan after he graduated from Princeton (or better yet, leaving out that Jeff’s mom’s father was ‘the regional director of the US’ Atomic Energy Agency in New Mexico… you know, the thing that managed Los Alamos) is certainly a choice.
Benjamin Domenech@bdomenech
Jeff Bezos was born in Alberquerque to a 17 year old high school mom who divorced his alcoholic dad before Jeff was 2. He worked at McDonald's before his grades earned his way into Princeton. He started Amazon in a garage with a 245k investment from his parents. That's reality.
English
can someone explain to me how this year's frontier AI models are *worse* than the average human?
to me, it seems they can reliably solve work most knowledge workers can solve, and ofc with way more breadth.
Chief Nerd@TheChiefNerd
🚨 Anthropic CEO Dario Amodei: “We are so close to these models reaching the level of human intelligence, and yet there doesn't seem to be a wider recognition in society of what's about to happen … There hasn't been a public awareness of the risks.”
English
@menhguin "we" meaning the market, chatgpt pro costs $200 a month, a human costs >$2000 a month. It doesn't matter than all the code is AI written or whatever, you can't run a company with just LLMs (for now). If you use AI on a daily basis you should know the limitations yourself.
English
@tenobrus I feel like everyone mentioned in this report came off like petty egomaniacal assholes (except Alec and maybe Daniela). I don't think the fact that Sam lies really matters, they all only care about achieving some selfish vision of the world, and have way too much power.
English
without further context, this kind of reporting could and perhaps should be easily written off as he-said-she-said, clash of big personalities, power struggles.
with further context, we know almost every member of leadership who remained at openai has since accused sam altman of consistent patterns of lying and manipulation. it is no longer easy or reasonable to dismiss this.
this is not particularly evidence that *Dario* is trustworthy. but taken as a whole it is quite strong evidence that Sam Altman is not an honest or good person, and we as the public should probably be quite unhappy he is in the position of power he is.
English
@redtachyon @paradoxenstein Can you point to a Veritasium video from over "a few years ago" that you liked? His content has always been popsci slop, but so is for example, 3b1b. Almost every other channel is less interesting or lower quality. I've been watching for 10+ years atp and it's just not worse.
English
This is really funny to me, because clearly I'm just part of his old audience (that's largely getting dissatisfied with his content), and you're part of the newer, wider audience that he's finding success in.
Obviously, he's very successful, especially in the last few years. But he achieved that by changing his content to be more suitable for people like you, not people like me.
I have a right to be a little bit salty.
English
This has got to be ragebait. Over the last few years, Veritasium has put out some of the best eduational content on all of youtube.
Ariel@redtachyon
Hopefully. You either die a hero, or do whatever this channel did over the last few years.
English
@HighFreqAsuka Also just seems like a complete nothing-burger college essay ass paper. What he learned was apparently just that it exists, and some of it is bad. You’d think if you go out of your way to write a paper on porn you’d have some interesting finding.
English
Absolutley insane tweet
Cas (Stephen Casper)@StephenLCasper
🧵🧵🧵 In the past few months, I have looked at hundreds, maybe thousands, of AI porn images/videos (for science). Here's what I learned from our investigation of over 50 platforms, sites, apps, Discords, etc., while writing this paper. papers.ssrn.com/sol3/papers.cf…
Türkçe
@peterwildeford OAI is probably benchmarkmaxxing this to keep the trend going. Claude, Gemini, and others are consistently below trend, despite claude being better/on par for agentic coding benchmarks. We’ll see for G3 though.
English
METR dropped this line in March and since then it's predicted the future beautifully
Daniel Eth (yes, Eth is my actual last name)@daniel_271828
Things are looking smoothly exponential for AI over the past several years, and I continue to think this is the best default assumption (until the AI R&D automation feedback loop eventually speeds everything up)
English
@tszzl Despite the US being filled with grifters, far more of its labor hours actually goes into producing real value. Countries like Korea, Japan, and to a lesser extent China are filled with hardworking people working on pointless BS. Inefficient businesses can be worse than fraud.
English
@johnloeber You’re hiring for a nameless company for god knows how little money, and you are still concerned about fucking version specific language knowledge about fucking python, what were you expecting?
English
The bar for technical interviews continues to fall lower and lower
I have taken to asking candidates what version of Python they write (so I can ask them some version-specific questions)
Then they tell me nonsensical version numbers, which is sometimes indicative that they're being aided by ChatGPT in the background
Had this today:
me: "what version of python are you writing these days?"
him: "3.5"
me: "are you sure? that came out 10 years ago"
him: "i, um, i don't think i'm prepared enough for this interview"
my jaw dropped in undergrad when i found out that this fizzbuzz test was real -- indeed, many candidates will not pass it. it turns out the bar is lower yet
English
@leothecurious Both these statements are completely consistent with what he said in the dwarkesh interview.
English
another one where karpathy seems to have outgrown his priors from a year ago:
"so, as an example, i think transformers are actually better than the human brain in a bunch of ways. i think they're actually an alot more efficient system, and the reason they don't work as well as the human brain is mostly a data issue, roughly speaking... And actually, as an example, like, transformers memorizing sequences is so much better than humans. like, if u give it a sequence and do a single forward-backward pass in that sequence, then if u give it the first few elements, it will complete the rest of the sequence. it memorized that sequence. and, it's so good at it. if u give a human a single presentation of a sequence, there's no way that u can remember that."
English
karpathy 1 yr ago, for perspective:
"the trajectories in ur brain when u're doing problem solving, if we had a billion of that, like AGI is here, roughly speaking, i mean, to a very large extent, uhm, and we just don't have that."
Dwarkesh Patel@dwarkesh_sp
The @karpathy interview 0:00:00 – AGI is still a decade away 0:30:33 – LLM cognitive deficits 0:40:53 – RL is terrible 0:50:26 – How do humans learn? 1:07:13 – AGI will blend into 2% GDP growth 1:18:24 – ASI 1:33:38 – Evolution of intelligence & culture 1:43:43 - Why self driving took so long 1:57:08 - Future of education Look up Dwarkesh Podcast on YouTube, Apple Podcasts, Spotify, etc. Enjoy!
English
"As ever - everything about how its trained could be improved dramatically. There is so much room to go. It is worth estimating how many similar jumps you expect over the next year."
This is my position on why fast progress is surely coming for a fair bit longer.
Sholto Douglas@_sholtodouglas
Claude 4.5 is the best coding model in the world - and the qualitative difference is quite eerie. I now trust it to run for much longer and to push back intelligently. As ever - everything about how its trained could be improved dramatically. There is so much room to go. It is worth estimating how many similar jumps you expect over the next year.
English
As a researcher at a frontier lab I’m often surprised by how unaware of current AI progress public discussions are.
I wrote a post to summarize studies of recent progress, and what we should expect in the next 1-2 years:
julian.ac/blog/2025/09/2…
English