Sabitlenmiş Tweet
Krish Ray
3.1K posts
Krish Ray
@KrishanuAR
Know thyself. Midwit.
Pittsburgh, PA Katılım Mart 2009
325 Takip Edilen146 Takipçiler
@KrishanuAR @elonmusk You are trying so damn hard not to say out loud what EVERYBODY really feels just not to piss the man off huh..
English
@AyakaMods Only explanation here is dude was drawing and hosting CSAM.
English
Google just permanently banned a manga artist’s entire Google account, just for uploading his own old manga files to Drive.
AI moderation triggered and flagged it, he tried to submit appeal then he got rejected it by Google and now he has lost everything like Gmail, Drive, all linked services is gone.
He never even sharing the files publicly, it’s only backing up his own a private work like any creator and artists.
This is Google Drive “AI moderation” in action. No human support and no serious to take action.
Physical storage or real private alternatives only.
Support the artists getting screwed by this. This level of corporate overreach is insane.
糸杉柾宏＠『寝取り魔法使いの冒険』第１第３ 月曜更新@masahiroitosugi
ところで恥を忍んで告白するのですが、私、Googleから垢BANされました。昔描いた漫画のデータをドライブにアップしている時に警告が出て、再審査請求も却下され、見事垢BAN。 まじで、困るよ。いろんなサイトやサービスにGoogleアカウントを使っていたので。 良い子のみんなには関係ないかもしれないけど、「俺、良い子…かな？」って人は気をつけてくれよな！
English
@NachoAMDG @AngloRestore Sure. The time the poster yearns for was when there was something of a caste system, which is ethically distasteful and does not exist de jure in Britain today.
English
@KrishanuAR @AngloRestore I said it was like a caste system, not the Hindu caste system
English
George Orwell was born and raised in India.
Is he Indian?
James Melville 🚜@JamesMelville
Aaron Rai was born and raised in England. He’s English.
English
@NachoAMDG @AngloRestore The British Raj did not fit into the Hindu caste system, so no that’s obviously not what I’m referring to.
English
Krish Ray retweetledi
Math is easy* because it has verifiable outputs and few messy judgement choices to make.
Which AI labs have the guts to make advancing social science a priority? It may actually do more for human flourishing to unlock sociology, econ & psych reseach.
* For AIs, not for humans
English
@Noahpinion Actual material concerns about AI aren’t held by the “anti-AI” crowd.
The people who seem to have the most well-thought-out and intellectually honest concerns about AI thus far have been those actually working on AI.
English
The people who think anti-AI sentiment is actually about water use are becoming more annoying than the people who actually think data centers use lots of water
Nick Gillespie@nickgillespie
Over at Substack, @JoshEakle asks: "It's 2026, and I have yet to see an anti-almond farm protest."
English
English
One of the reasons homeschooled kids have superior educational outcomes is avoiding the slow-progress-across-all-subjects method public schools impose on every student, no matter how they learn.
The evaluation/testing you are talking about would almost certainly prohibit that sort of tailored education, especially since they would be designed and administered by a system that wants to eliminate homeschooling in almost all cases.
English
Right, if homeschooling is actually super high quality, then homeschooling families should not object to being evaluated, tested, and checked-in-on to make sure their kids are actually learning.
𝙷𝚒𝚌𝚔𝚖𝚊𝚗@shagbark_hick
Look, pro-homeschooling people are piling on this guy but he's not wrong at all. I know of families who are straight-up failing to teach their children beyond a 2nd-grade education. Like 16-year-old kids who don't know what the Capital of the US is, have never heard of Toronto, and could not perform a basic algebra problem. While it's true that there are many excellent homeschooling families whose instruction far exceeds that of any public school, the fact is that the American system of "school your kids however you want" does also yield a lot of severely stunted kids who get stuck playing "catch-up" for the rest of their lives. My wife was homeschooled, we might homeschool, I'm not averse to the practice per se, but it's a fact that there's a whole world of multi-generational homeschoolers who are not and will not rise above a 2nd or 3rd grade education level by their own volition. Should the state intervene in these cases? I dunno, the answer to the question is not as "easy" as many make it out to be. The idea of barely getting through 'Mary Had a Little Lamb' at the age of 16 should be appalling, and it IS happening in some American homeschooler families.
English
@DarrigoMelanie If only revenue flowed to programs so cleanly, without being devoured by thousands of other interests along the way.
English
Jeff Bezos paid $500M for his super-yacht and $75M for his super-yacht’s mini-yacht — both of which he’s allowed to write-off on his taxes.
That alone would cover $180 in classroom supplies for every public school teacher in the U.S.
Aaron Rupar@atrupar
Bezos on CNBC: "You could double the taxes I pay, and it's not gonna help that teacher in Queens. I promise you."
English
@vijaythirumalai A lot of his friends probably work at Anthropic, and they lured him in by talk of all the advanced work they were doing but couldn’t share unless he joined.
English
I simply cannot fathom this move
Dario doesn’t look that charismatic to me atleast from his interviews
Andrej could have joined
1/ Elon
2/ Zuck
3/ Google
4/ Apple
5/ MSFT
Everyone would have given him a blank check and much bigger role and platform
How was Dario able to hire him ?
Crypto India@CryptooIndia
BREAKING: OpenAI co-founder Andrej Karpathy joins Anthropic.
English
@yacineMTB It’s a Google model though. It’s gonna be way worse in practice.
English
@Noahpinion It kinda sucks… Even if that's what they truly believe and they wanted to work openly to solve that potential outcome, this suggests a strong incentive for them to lie in the future.
English
I guess AI leaders constantly yelling that their goal was to put everyone on welfare for all eternity didn't win many friends.
Semafor@semafor
AI backlash is growing in the US, especially among young people. Polls show that 70% of Americans think AI is moving too fast, over 50% have negative views of it, and just 18% of young people say they feel hopeful about it. semafor.com/article/05/18/…
English
@MrTimeAttack @elonmusk @UN_Women You're stating a not-universally-accepted paradigm and nomenclature that was proposed only a bit over 30 years ago as an obvious and immutable Truth.
Such zealous religiosity can never lead to intelligent discourse or realistic progress.
English
@GaryMarcus Ideas are cheap, has the doppelgänger actually produced supporting work, or is it just ideas?
English
seriou question: how do you handle an intellectual doppelganger who has systematically started adopting every position you have argued for for 30 years while presenting each idea as if it were his own?
nothing in graduate school prepared me for this.
English
@GaryMarcus @RichardSSutton @HamidMaei Yes, but the next generation of models will be pre-trained with the context that those 50 tools provide, rendering them obsolete. So that human knowledge will not be relevant to provide to the model anymore.
English
i wonder whether this needs an update.
current methods, such as they are, leverage massive amounts of human knowledge as their primary fuel. they would be lost without it.
and they even build some knowledge into their system prompts.
and lately they build knowledge into their harnesses, usually by over 50 tools that have been carefully crafted with human knowledge.
English
The bitter lesson in 26 words:
Don’t be distracted by human knowledge, as AI has been historically.
Instead focus on methods for creating knowledge that scale with computation, like search and learning.
English
Been using Claude as an architect/designer and codex as the implementer for a home IoT project, and incidentally this is essentially the pattern that Claude came up with.
Codex is the one that keeps the running implementation notes, which Claude reads between phases and makes adjustments to upcoming implementation phase goals.
English
a prompt I've been using a lot recently:
implement <SPEC> and while you do, keep a running implementation-notes.html file (or markdown) with decisions you had to make weren't in the spec, things you had to change, tradeoffs you had to make or anything else I should know
English
One thing to watch for with Claude & GPT is that the models expose too much irrelevant history in their outputs. Slides are given footers saying things like "Better, more targeted version" if you asked for a better version, documents make references to how they are improved, etc
English
