Modeling societal-scale deliberation, developing decision-making models, building libraries of knowledge, and offering educational content on complex issues.
Retweets ≠ endorsements.
Joined August 2012
- Tweets1.7K
- Following2.5K
- Followers6.4K
- Likes1.5K
The Society Library 📖 retweeted
A single debate about one nuclear power plant generated 5,862 distinct arguments. The human mind holds 4 to 7 things at once.
@JustJamieJoyce calls this the core problem in democratic governance, and thinks AI might be the way to close the gap. 🧠
youtube.com/watch?v=Ep6GdRS7…
The Society Library 📖 retweeted
I was sent @slatestarcodex's article on "Your Attempt To Solve Debate Will Not Work"
So I wrote a quick reply: "Your Attempt To Refute Argument Maps Will Not Work" in the voice of Scott Alexander.
justjamiejoyce.substack.com/…
Felt cute, may make a debate map about it later.
The Society Library 📖 retweeted
I was not previously aware that the @SocietyLibrary was mentioned in a paper by @AnthropicAI researchers and the @compdem project.
Thanks for the nod, folks.
arxiv.org/abs/2306.11932
The Society Library 📖 retweeted
Really appreciate Jamie Joyce’s thoughts on controversial figures as information sources, I’m more apt to trust the structure of SL knowing someone who values truth over character derangement syndromes is at the helm.
The Society Library 📖 retweeted
Join us on Wednesday July 2 at 12 PM ET for Metagov Seminar: @JustJamieJoyce will present on her work with @SocietyLibrary
Learn more and register here - lu.ma/bsqnutec
The Society Library 📖 retweeted
Thank you for having me on your podcast, @Liv_Boeree! It was such a pleasure. You are an incredible host.
Also - beyond making sense of people's values, we're building out structured representations of logical arguments and the evidence which backs them.
It's important to detect and incorporate predominant value preferences of a given civilization, but we also must gather the collective evidence and logical reasoning which a given civilization generates. Its about harnessing the best collective intelligence we have to manage ourselves well.
Really appreciate you giving me an opportunity to discuss this at length!
Democracy is a hot mess right now, in part because politicians are an outdated technology. What if we replaced - or at least supplemented them with -databases that tracked their constituents' values?
That's one of many juicy topics covered in this Win-Win conversation with @JustJamieJoyce . Jamie is the founder of Society Library and a leading expert on information infrastructure, and this is one of my favourite conversations thus far.
Watch right here on X, or your usual pod platform.
Chapters:
0:00 - Intro
4:30 - Diablo Canyon Debate Map
14:26 - How To Evaluate Evidence?
16:18 - How To Minimize Bias?
19:06 - Bad Faith & Fringe Perspectives
24:16 - Flat Earth Conspiracies?
28:37 - Cyber Attacks On Internet Archive
33:54 - Modernizing Government
41:01 - The Evil of Corporate Lobbying
49:40 - Managing Info Hazards
53:19 - Virtues And Values
57:06 - When Jamie Neatly Went Blind
1:01:27 - Improving The Information Ecosystem
1:12:44 - Prediction Markets and Sensemaking
1:15:25 - Structuring Data For Improved Learning
1:22:17 - Society Library Meme Token Nonsense
1:32:21 - Rapid Fire Predictions
1:38:51 - Jamie's Favorite Win-Win
The Society Library 📖 retweeted
Spicy take: politicians are an outdated technology that should be replaced by viewpoint databases.
@JustJamieJoyce explains why 👇
The Society Library 📖 retweeted
I purchased OpenAI's Deep Research tier, and honestly, I still prefer the @museofresearch to retrieve quality information on topics I am researching. I am so over prompting and arguing with LLM systems. The @SocietyLibrary's Muse of Research has quality built-in and its free. That's how it should be.
The Society Library 📖 retweeted
Work we are doing on the @MuseofTruth:
- Working on her over-responses (if replied to, she responds too much, she's going to do it to this tweet, lol)
- Building another layer to her pipeline so she optimizes for only the most appropriate forms of content (e.g. scholarly works for scholarly issues, economic docs for economic issues, etc.). She does pretty great, but she can be better.
- See what we can do about temporality (this is a tricky one). We saw one instance of her pulling outdated information.
Otherwise, thank you ALL so much for using her and helping us find these edge cases. We can't wait to make her better! 🫶
The Society Library 📖 retweeted
Wrote an op ed today following the release of @museoftruth and @museofresearch.
Let’s reclaim agency, one tool at a time.
societylibrary.medium.com/ca…
The Society Library 📖 retweeted
platforms like @SocietyLibrary are trying to bring transparency and truth into the 'code' of knowledge – it's all about connecting the dots and seeing the bigger picture
The Society Library 📖 retweeted
'The Difference Between Prediction Markets and Debate (Argument) Maps.' I wrote this quickly today because it seems to come up not infrequently: lesswrong.com/posts/zi3WW3ow…
The Society Library 📖 retweeted
Re. @Meta and Mark Zuckerberg's announcement on replacing fact-checkers with community notes. First, I love the concept of community notes and recognize that fact-checkers can have political bias (I've written about this myself)...however...the work of fact-checkers should not be discounted.
Many fact-checkers are investigative journalists who pick up the phone, call people, and write emails to interrogate the truth out of individuals, companies, etc.. They dig deep into details and perform real investigative work as their full-time job. While community notes can help debunk more easily verifiable inaccuracies, until everyday users are willing to put in a lot of legwork, some deeper investigative services may be lost. And as far as I know, AI is not capable of this kind of work yet.
The Society Library 📖 retweeted
Prediction markets help answer the question "will X happen?"
Debate maps help answer the question "should X happen?"
The Society Library 📖 retweeted
An experiment in generosity: Charity backed by ai16z’s @shawmakesmagic receives 500 million SL token donation
@JustJamieJoyce affirmed that @SocietyLibrary has embraced SL as a community token and the charity will continue holding
📜: blocmates.com/news-posts/cha…
The Society Library 📖 retweeted
gm everyone! ☀️ Well, the mystery of who created $SL coin is still a mystery, but the @SocietyLibrary owns half, and we are exploring locking it to protect this community. More to come soon as we quickly learn best options, but for now - we are excited to be in this with you. The power to seek the truth should be in everyone's hands - and if we have anything to say about it as a charity dedicated to that mission for years, that's what this coin will be about. Thank you all *so much.* 🙏
The Society Library 📖 retweeted
Love these experiments.
Non profit AI project ongoing.
Someone creates a token and donates 50% of supply.
Founder of project pledges to hold.
Result = low float token.
Sure, they'll eventually sell, but it'd be churlish to dump straight away - at least that's the bet.
This evening we were surprised when someone created $SL coin and donated us half of it. We are currently holding and are excited to be welcomed into the crypto community. For those who are just meeting us, here's a podcast about our mission: podcasters.spotify.com/pod/s… (pre LLM boom)
The Society Library 📖 retweeted
Thank you all for joining the @SocietyLibrary AMA! This is an incredible community and we appreciate your support! *Please* ask me about our tech. Off to that dinner I told y'all I delayed to chat with you all. Will BRB 🫶
The Society Library 📖 retweeted
Reducing Intelligence Asymmetry / Scaling inquiry into Truth 🔍
Our work automating our methods to identify experts, search and retrieve relevant content, and ground graphs to map scientific fields + policy debates.
The Society Library 📖 retweeted
When it comes to truth-seeking, the fidelity of claims and data matters. Getting folks to critically examine previously unexamined beliefs to get them to change their minds has been a hand-holding technique known for a while. With AI, that could scale, however.
The now named "debunk bot" was featured in the NYT, and it was instructed to "very effectively persuade" people who hold conspiracy theory beliefs, or beliefs close to a set of pre-defined conspiracy theory statements, to doubt and disbelieve them. Apparently, they did this successfully via three rounds of dialogue with an LLM which took, on average, 8.4 minutes.
But what else was observed? In the NYT article, the published article is referenced. However, there is more detail in the preprint. Let's dive in a bit.
➡️Caveat, caveat: this is my individual reading on a Friday afternoon, and I am a busy & fallible human who works on deliberation/collective intelligence for truth-seeking techniques for a reason. All corrections to my mistakes via the hive mind will be greatly appreciated.
NYT Article: nytimes.com/2024/09/12/healt…
Science Article: science.org/doi/10.1126/scie…
PrePrint: osf.io/preprints/psyarxiv/xc…
