고정된 트윗
Scipy Japan 2019 schedule is on: talks about chainer, tensorflow, RAPIDS and more: scipyjapan2019.scipy.org/schedule all the talks will be available in both english and Japanese through translation
English
David Cournapeau
30.1K posts
David Cournapeau
@cournape
Amazon, ex-search/recommendation/ML at @mercari_jp. Former NumPy/Scipy core contributor. Occasional musings on economics/music/Japanese culture
Tokyo, Japan 가입일 Haziran 2010
1K 팔로잉1.7K 팔로워
David Cournapeau 리트윗함
David Cournapeau 리트윗함
Behold our GOLF COURSE and tremble, world!
The White House@WhiteHouse
AMERICAN EXCEPTIONALISM 🇺🇸
English
David Cournapeau 리트윗함
TIM DILLON: We’re not even winning the trash talk war. You’d think America would win that at least, if we’re going to win one thing. Instead, we’re getting bodied by Iranian AI in the war of trash talk. How embarrassing. We invented this, and we’re getting lit up.
😂😂😂
English
@brianluidog And then you go into a small shop that sells spicy ramen that will make you bleed. Very hard to guess before trying
English
David Cournapeau 리트윗함
If there’s one thing I know about PhDs, it’s that they are the best negotiators and very grounded realistic policy makers.
English
David Cournapeau 리트윗함
What if Trump were really playing 4D chess, but instead of being a China hawk as Hudson Institute types assume, he was actually Lemoine-pilled on the idiocy of geopolitics and was covertly working to make Chinese regional hegemony so inevitable that it's pointless for the US to even try to prevent it, thereby saving the world from a conflict that would be not only pointless but also incredibly wasteful?
Geoff Wade@geoff_p_wade
"The U.S. Is Pushing Southeast Asia Toward China. The Iran War Made It Worse." by Joshua Kurlantzick cfr.org/experts/joshua…
English
David Cournapeau 리트윗함
Jews, famously hateful of preserving written text.
barney@barneyxbt
just a reminder that 95% of the worlds history is unknown due to the burning of the library of alexandria.. guess who burned it
English
David Cournapeau 리트윗함
David Cournapeau 리트윗함
C’est Kévin qui les perçoit.
Subventionner la demande sur un marché inélastique, c’est juste gonfler les prix.
Estelle Midi@EstelleMidi
💶 Impôts ? 📞 Kévin : "Si les Français veulent être patriotes, qu'ils mettent la main à la poche ! J'ai eu un locataire, en âge de travailler, sur un loyer de 900€ il perçoit 536€ d'APL ! On va me prendre des impôts sur mon revenu locatif pour aller verser 536€ d'APL !"
Français
David Cournapeau 리트윗함
David Cournapeau 리트윗함
Mishima Yukio? The UFO researcher?
eric ゑリｯ久@shinobu_books
Mishima Yukio was a member of the Japan Flying Saucer Research Society日本空飛ぶ円盤研究会議 He even submitted a letter about watching for UFOs over Atami: Every night, every night I hoped that maybe tonight a UFO would land, but alas I was never able to witness such an event.
English
@brianluidog @AndyMasley Yeah, there is a bit of lack of perspective from tech playing the victim, when you see how a certain part of tech reacted to Biden because of raising taxes.
English
David Cournapeau 리트윗함
@AndyMasley This is true, but it's because cars and big ag didn't upend the journalism business model.
English
A sizable number of journalists covering tech have a very naive interpretation of what it means to "speak truth to power" where they're hyper selective of which powers they speak truth to at the expense of letting other often worse powers off the hook. Like cars or big ag
Timothy B. Lee@binarybits
"Among my tribe, journalists, mostly on the east coast, the acceptable stance is mainly just: adversarial, always."
English
David Cournapeau 리트윗함
@Austen The Internet did not eliminate middlemen and we're 30 years in. It did however create new middles and expanding both sides of their markets.
English
@finbarrtimbers I also believe it was key to make python popular for science because multithreading at the c boundary was opt in. A lot of scientific code was just not thread safe (or even re entrant), especially Fortran.
English
The GIL primarily exists to make CPython’s memory management thread-safe. It has also become a load-bearing part of Python's API, as, in a classic example of Hyrum's Law (hyrumslaw.com), much Python code relies on the thread safety that it has created.
English
I wrote about the GIL, a source of perennial confusion in high-performance Python, as I found myself confused by how, exactly, it works.
finbarr.ca/what-is-the-gi…
English
David Cournapeau 리트윗함
As AI reasoning gets good enough, we think memory will be the next bottleneck for agents. Can your agent improve with more experience?
We call this Memory Scaling, and it's related but different from continual learning. A few examples and challenges:
databricks.com/blog/memory-sc…
English
David Cournapeau 리트윗함
Passionnant essai, un peu pessimiste, certes, mais comme toujours chez @brunopatino c'est un pessimisme stimulant, jamais rabougri, en sollicitant maints exemples du passé (remarquable chapitre sur Gutenberg) et du présent. D'une intelligence pas du tout artificielle, donc.
Français
David Cournapeau 리트윗함
Anthropic killed this, Anthropic killed that, why cant Anthropic kill TurboTax
English
David Cournapeau 리트윗함
French gov moving faster than most companies
we now have a skill to do taxes, what a time to be alive
arthaud@arthaud_
the French government’s MCP is better designed than 99% of MCP servers coming from tech companies citizens can use agents to understand how their money is spent
English
David Cournapeau 리트윗함
i co-wrote the Anthropic engineering blog on Claude Managed Agents, and wanted to share some thoughts on agent harnesses + infrastructure for long-horizon tasks ... 🧵
anthropic.com/engineering/ma…
English