e
“I just build notice explainer ai using @lovable_dev I built this to make sense of confusing tax notices that needs a clear breakdown. You can try it free with 5 credits, and I’d love your feedback! Also, feel free to test it on any kind of document—not just tax notices.” / X https://x.com/abhirb/status/1900173948794409074
“I regret to announce that the meme Turing Test has been passed. LLMs produce funnier memes than the average human, as judged by humans. Humans working with AI get no boost (a finding that is coming up often in AI-creativity work) The best human memers still beat AI, however. https://x.com/emollick/status/1901431681279475808
“So far, as LLM model size gets larger, it seems to have a direct effect on reducing known problems with LLMs. Bigger LLMs hallucinate less, show less bias, and are less sensitive to prompting style, among other things. Not that these problems go away, but they do decrease.” / X https://x.com/emollick/status/1901776990962630666
“Thinking from first principles instead of pattern-matching the past lets you anticipate important changes with a bit of advance notice. Not all of them, but some big ones.” / X https://x.com/fchollet/status/1902477232029000008
Hollywood’s response to AI Action Plan https://s3.documentcloud.org/documents/25590860/hollywoods-response-to-ai-action-plan.pdf
“Exciting job I’m hiring for – Policy Demos! We’ve often found the best way to help people understand powerful AI technology is to ‘show, not tell’, and the best way to do this is to demonstrate the real capabilities of real systems. Help us do that! https://x.com/jackclarkSF/status/1901789490437669370
“@willccbb My cynical take is that: to the extent that these are better, they’re not really LLMs anymore. Folks couldn’t build better LLMs, so they started building LLM systems (CoT) and shifted the goal posts accordingly. The community could always build those, though not as scaled. The” / X https://x.com/lateinteraction/status/1901642081770295732
“There was controversy over the way the big AI companies did image generation a year ago. Now Grok can make pictures of famous people, Gemini can remove people or watermarks from photos, etc. Is it a shift in responsibility from firms to users? A lack of negative impact? Apathy?” / X https://x.com/emollick/status/1901798161846595851
“New skill unlocked: Gemini 2 Flash model is really awesome at removing watermarks in images! https://x.com/deedydas/status/1901042632958345369
The court rejects Elon’s latest attempt to slow OpenAI down | OpenAI https://openai.com/index/court-rejects-elon/
“”This is the year that AI gets better than humans at programming forever. And there’s no going back.” OpenAI CPO Kevin Weil highlights the rapid improvement of AI models in competitive coding, predicting one will reach the number 1 spot in 2025. He says AI surpassing humans in https://x.com/vitrupo/status/1901258363364897016
“I suspect that a lot of “AI training” in companies and schools has become obsolete in the last few months As models get larger, the prompting tricks that used to be useful are no longer good; reasoners don’t play well with Chain-of-Thought; hallucination rates have dropped, etc.” / X https://x.com/emollick/status/1901011090068197484
“I wrote a quick new post on “Digital Hygiene”. Basically there are some no-brainer decisions you can make in your life to dramatically improve the privacy and security of your computing and this post goes over some of them. Blog post link in the reply, but copy pasting below https://x.com/karpathy/status/1902046003567718810
“id like to publicly state i dont buy the whole ‘there will be a ton of new jobs’ thing for normal people. there will be many new jobs but not for normal people” / X https://x.com/nearcyan/status/1901932030386127224
“We’re partnering with @Kalshi for March Madness to give you matchup predictions and odds for NCAA basketball. We will be doing more things with them to surface odds and predictions for live events around finance, crypto, politics and sports.” / X https://x.com/AravSrinivas/status/1902044102059028575
“Perplexity released DeepSeek-R1 1776, an updated version of a model originally developed for China. This new version removes the political censorship built into the original, making it more useful outside China. The fine-tuning process involved removing blocks on politically https://x.com/DeepLearningAI/status/1902710466662523099
“Introducing: ShieldGemma 2 – a 4B model for image safety classification 🛡️ 👀Use as input filter for VLMs ❌or for blocking dangerous image generation outputs 🦺Great for production deployments Blog: https://x.com/osanseviero/status/1901764379328037047
“one of the greatest Chinese advantages that I notice, and that gets little attention, is how much less afraid their boomers are of learning 2 tech. The state machine jumping on R1 adoption is illustrative. It implies they’re not so limited by population aging as often argued.” / X https://x.com/teortaxesTex/status/1902545539725463758
