It leaves me wishing we had access to models that were blank slates, maybe not entirely blank so as not to be practical but blank in terms of unbiased by product outcomes the providers wish to have.
English
Nathan Borror
8.3K posts
Nathan Borror
@nathanborror
Do not be afraid
California, 🇺🇸 Katılım Kasım 2006
531 Takip Edilen2.3K Takipçiler
At the risk of sounding behind, I found this interview with Peter Steinberger[1] and this LessWrong[2] post illuminating. Before I naively understood LLMs as needing instruction but that's not necessarily true, they need to be primed. It's more like putting your thumb on the scale in the right parts to create a trajectory towards an specific outcome.
1. youtube.com/watch?v=YFjfBk…
2. lesswrong.com/posts/vpNG99Gh…
YouTube
English
This is fun and seems difficult to block. The output isn't strong but the scraping mechanics work. Tried parallelizing the browser sessions but everything slowed down too much and the automated scrolling didn't work as well.
English
@elonmusk Any way you can do a deal with Anthropic so Claude memories can be accessed by Grok?
English
I wish X gave me an option to make a clip from a video URL (mostly YouTube) so I didn't have to use yt-dlp and ffmpeg to make one myself.
English
Noticing this when I try to programmatically scroll the page to capture content as it loads. Works fine in a WebView but not in a standalone WebPage.
English
Anyone know how WebPage changes when it’s placed inside a WebView? It seems as though the underlying page javascript can detect when it’s headless and it’d be nice if there was a way to obfuscate this.
Nathan Borror@nathanborror
Hard to believe Apple subtly shipped a very nice headless browser—WebPage—and few are talking about it.
English
Begs the question, what did he see?
Andrej Karpathy@karpathy
Personal update: I've joined Anthropic. I think the next few years at the frontier of LLMs will be especially formative. I am very excited to join the team here and get back to R&D. I remain deeply passionate about education and plan to resume my work on it in time.
English
Two Popes was a good movie. I hope Leo's technical counsel on ai is wider than a cofounder.
Vatican News@VaticanNews
Pope Leo XIV’s first encyclical, Magnifica humanitas, on preserving the human person in the age of artificial intelligence, will be released on May 25. A presentation event with the Pope and various speakers is scheduled for the same day at the Vatican. vaticannews.va/en/pope/news/2…
English
“The people who are building real companies are in Palo Alto.”
People are saying…
Kai@kai_xbt
Avi Patel says nobody builds real companies in SF anymore "SF is filled with striver people that only care about fundraising and not about building real businesses. All they go out and do is brag about their last valuation and then you look at their company and it's a fucking shitshow. No one knows how to build real shit in SF. I'm dead serious." "It's not what it used to be at all. Anybody that's in SF right now doesn't like it. The people who are building real companies are in Palo Alto. All the ones founding companies in SF are like striver Gen Z idiots who don't know what the fuck they're doing, have never had a girlfriend, never had a real interaction with a human being and don't have any friends."
English
True and too often employees at large companies use "safety" as a distraction to slow down progress that may affect their job.
Ashlee Vance@ashleevance
Alex Wang on the problems he found inside of Meta. #1, he says, was that the company lacked the AI religion of other frontier labs.
English
First it was function calling and structured outputs, then tool calling in a loop, then reasoning in the loop, then adding memory, then SKILLs to manage context window, then “dreaming” to manage memory, then harnesses to bundle everything up with a SOUL… what next?
English
Someone needs to write The Art of the Agent Harness. Maybe an agent?
English