Are there any aspects you dislike about it?
Posts by Changbai
It's crazy and terrible to eliminate SBE, and especially the idea NSF would do it in violation of Congressional direction.
Even if you only cared about AI, it's important. SBE's support of opinion data has crucially helped AI alignment research; for example, data from GSS for LLM eval & modeling.
This is very interesting paper, thanks for sharing!
Can I understand the finding as "The problem is tokenizers + length of the word"?
what are we looking at
This article has been making the rounds, but someone else pointed out that the "evidence" is based on LLM-simulated users. I have extremely low faith in the validity of these results, especially given the established stickiness of political beliefs and the known issues with in-silica simulation
Carlini is an incredible scholar, well-known before joining Anthropic. I think his opinions are very worth listening to.
🫡
fascinating new dimensions of brainrot, I could never. this guy's entire account is "finding <object in space> from <reference location> using <this visualiser>"
seems like many products have attempted to solve this problem. have you looked at the other ones?
uhhhhhh
poster for what 👀
Aaron Swartz
never forget, never forgive
otherwise big fan of 100 rabbits, they're awesome
@anildash.com also comes to mind
@pvh.ca is just a lovely human
@bmann.ca may not be a dev, but inspirational nonetheless
but what if claude write s the configs, think of all the possibilities
Thanks! Curious stuff. Which class introduces them? - or is it something encountered outside the classes?
AIC and BIC?
ChatGPT: you’re so right, Leto Atreides II. Merging with a worm would give you unconscionable power that you, alone, could wield — you, alone, could be trusted with. You wouldn’t be committing a genocide. You would be saving humanity
The first message on ARPANET was sent between UCLA and Douglas Engelbart's Augmentation Research Center. Pioneers of the internet certainly had intelligence augmentation on their mind.
This is insane!
source?
imagine a prompt injection
Thanks for posting this. The stigmatization of LLM is at this point doing more harm than good. It's frustrating that some can't recognize that technology is not their creators, and their impact depends on the users.
Amazing analogy, please include in the course
LLMs are piñatas.
You take a bunch of newspapers (web corpus), cut and mix them all up (stochastic gradient), form them into paper mache that looks anthropomorphic (RLHF), and finally attach a harness (Claude Code) so everyone can get candy.
I like candy. I also eat vegetables. You can do both!
Hmm I just realized that when I first read the tweet I thought word processors refer to tools like Grammarly... oops. A google doc with no AI suggestion is closer to a "faster pen& paper" and doesn't have significant semantic & syntactical influence, so I think your comparison was fair.
I remember seeing one research that looked into the difference of memorization between note-taking with pen & paper, v.s. with keyboard. The main factor that created the effect was speed of input. So speed def affects our thinking.
I don't think this is a fair comparison? Word processors differ in that they manipulate at the semantic and syntactical level, whereas writing tools differ only in the speed and locomotion of input.
FWIW, word processors also affect a writer’s voice, which is why some authors prefer to do the first draft with pen on paper.
If you use AI to help you write, it affects what you say and how you say it. It’s a powerful tool. Use it with intention.
IP is not a new debate.
Greek gods were all like, “YoU WoUlDnT DoWnLoAd A FiRE” and Prometheus was all like “watch me!” and then a giant eagle/legal was like, “you must pay $100 million or your liver.”
Does anyone know of any recent studies around the effectiveness of different extrinsic (e.g. financial) vs intrinsic (e.g. community altruism) motivators for open source contributors?
Asking for me, a friend, and also the rest of the OSS community.
and this thing specifically!