A couple of weeks ago, AI company Anthropic published the constitution that they use to train their Claude LLM (“under a Creative Commons CC0 1.0 Deed, meaning it can be freely used by anyone for any purpose without asking for permission”). From the company’s news release:
We’re publishing a new constitution for our AI model, Claude. It’s a detailed description of Anthropic’s vision for Claude’s values and behavior; a holistic document that explains the context in which Claude operates and the kind of entity we would like Claude to be.
The constitution is a crucial part of our model training process, and its content directly shapes Claude’s behavior. Training models is a difficult task, and Claude’s outputs might not always adhere to the constitution’s ideals. But we think that the way the new constitution is written — with a thorough explanation of our intentions and the reasons behind them — makes it more likely to cultivate good values during training.
The full document is 80+ pages, but the news release does a decent job in summarizing what’s in it.
Claude’s constitution is the foundational document that both expresses and shapes who Claude is. It contains detailed explanations of the values we would like Claude to embody and the reasons why. In it, we explain what we think it means for Claude to be helpful while remaining broadly safe, ethical, and compliant with our guidelines. The constitution gives Claude information about its situation and offers advice for how to deal with difficult situations and tradeoffs, like balancing honesty with compassion and the protection of sensitive information. Although it might sound surprising, the constitution is written primarily for Claude. It is intended to give Claude the knowledge and understanding it needs to act well in the world.
We treat the constitution as the final authority on how we want Claude to be and to behave — that is, any other training or instruction given to Claude should be consistent with both its letter and its underlying spirit. This makes publishing the constitution particularly important from a transparency perspective: it lets people understand which of Claude’s behaviors are intended versus unintended, to make informed choices, and to provide useful feedback. We think transparency of this kind will become ever more important as AIs start to exert more influence in society.
Casey Newton and Kevin Roose recently interviewed the primary author of the constitution, philosopher Amanda Askell, for the Hard Fork podcast (the segment starts at ~25min).
Newton says the document reads like “a letter from a parent to a child maybe who’s leaving for college”:
And it’s like, we hope that you take with you the values that you grew up with. And we know we’re not going to be there to help you through every little thing, but we trust you. And good luck.
Both the constitution and the conversation with Askell are fascinating, no matter where you lie on the AI debate continuum. You might also be interested in this video of Askell answering questions from Claude users about her work:
Every time there’s a new social media app or network that breaks out, someone writes an article about how this new network encourages people to be themselves and have fun without all of the heaviness of other platforms. The latest example of this is Kevin Roose’s NY Times piece about TikTok.
TikTok has none of that. Instead, it’s that rarest of internet creatures: a place where people can let down their guards, act silly with their friends and sample the fruits of human creativity without being barraged by abusive trolls or algorithmically amplified misinformation. It’s a throwback to a time before the commercialization of internet influence, when web culture consisted mainly of harmless weirdos trying to make each other laugh.
In 2016, Jenna Wortham wrote this about Snapchat for the NY Times:
Its entire aesthetic flies in the face of how most people behave on Facebook, Instagram and Twitter — as if we’re waiting to be plucked from obscurity by a talent agent or model scout. But Snapchat isn’t the place where you go to be pretty. It’s the place where you go to be yourself, and that is made easy thanks to the app’s inbuilt ephemerality.
In 2013, Mat Honan wrote this about Vine:
It built a ground up culture that feels loose, informal, and — frankly — really fucking weird. Moreover, most of what you see there feels very of-the-moment. Sure, there’s plenty of artistry that goes into making six second loops, and there are volumes of videos with high production values. But far more common are Vines that serve as windows into what people are doing right now.
Implicit in these pieces is the idea that there’s something intrinsic to these apps/networks that makes them hew closer to real life and/or lightheartedness than older and bigger platforms…the ephemerality of Snapchat, the ease of shooting a Vine video, the fun filters and templates of TikTok. Some part of that is surely true, but what if being small and new is the thing that makes these networks fun? As I wrote in response to Wortham’s article a couple of years ago:
Blogs, Flickr, Twitter, Vine, and Instagram all started off as places to be yourself, but as they became more mainstream and their communities developed behavioral norms, the output became more crafted and refined. Users flooded in and optimized for what worked best on each platform. Blogs became more newsy and less personal, Flickr shifted toward professional-style photography, Vine got funnier, and Twitter’s users turned toward carefully crafted cultural commentary and link sharing. Editing worked its way in between the making and sharing steps.
TikTok probably feels a lot like Flickr or Twitter in the early days, where everyone is exploring and the users are all kind of doing the same things with it. As networks get bigger, they reach a point where there isn’t just one big group exploring the same space together. Instead, you have many big groups who have different goals and desires that all need to fit under one roof (essentially, politics becomes necessary)…and that can get messy, particularly when the companies running these apps want to appeal to the widest possible audience for capitalization purposes.
Novelty is probably the biggest factor though. TikTok is fun because it’s new. When you join up, you get new superpowers and flexing those abilities gives the old brain a shot of dopamine, particularly when the flexing is social. Later, when many of the social possibilities have been explored and even exploited, fun becomes harder to come by. Even Twitter can still be fun — see the replies to Wortham’s recent tweet about fave NYC moments — but the templates for interaction on the platform have long since been set in stone. It would be very surprising if a large & mature social network came along that didn’t also get less fun and “real” as it developed. That would be a special achievement.
Cronuts are donuts made from croissant dough and they are all the rage here in NYC. They were invented by chef Dominique Ansel and they are only available in limited quantities at his bakery in Soho. Apparently people start lining up for them at 6am and all 200 of the world’s daily supply of cronuts are gone within minutes of opening. Naturally, a black market has sprung up, with cronuts selling on Craigslist for upwards of $25/item:
Kevin Roose has some ideas for Ansel about expanding the reach of the cronut, but in the meantime, Edd Kimber replicated the treat at home with a quickie croissant dough.
Since I wont be in New York any time soon I thought I would see if I could replicate them at home, and you know what? They are pretty damn good! Now the dough I’m using isnt a proper croissant dough, its my quick dough made with just 20 minutes active work which, compared to traditional croissant dough is a snap to make.
Update: Pillsbury has gotten into the act as well with a cronut recipe that uses their crescent dough.
Update: In his forthcoming cookbook, Ansel reveals the recipe for the cronut and it’s not for the faint of heart…the recipe takes three days to make.
Socials & More