decentralized and democratic differential defensive acceleration・ https://vitalik.eth.limo/general/2023/11/27/techno_optimism.html#dacc “My techno-optimism” · https://vitalik.eth.limo/general/2025/01/05/dacc2.html “d/acc: one year later”

Main

Read article →

d/acc: one year later

d/acc

d/acc: one year later

https://vitalik.eth.limo/general/2025/01/05/dacc2.html

There’s a guy on twitter spending his whole day hate replying to every single “Free Roman Storm" tweet

What a loser

So far, the judicial system has stayed truest to American values, even as the other branches have drifted further from them

If Roman Storm isn’t freed, it will be a dark sign of what’s to come

there are lots of ways to go about it and this is not my area of expertise, but,

I do like what I’ve seen of SYCON-Bench, so it would probably be based on that. I think for reasoning & instruct-tuned models, there’s probably a relationship where the drop-off in needle-in-a-haystack tests correlates with an increased tendency to agree with the user.

So I’d actually be more interested in running the tests twice, one at the beginning of the context window and one after at least 16K words of whatever random chat history that contains the user’s preferences.

Side note: I’m pretty sure that infamous GPT-4o sycophancy rollback was at least half-due to weaknesses that become very noticeable amid long context (which would include Memory and “Reference Chat History” types of features).

For the benchmark, I would also want some kind of module that has prompts to test the LLM’s psychotic break/mental health emergency responses, like if a user suddenly decides they’re Jesus or that the LLM is a god or they’ve discovered a global conspiracy.

A related effort would be a set of prompts that tests the LLM’s ability to discriminate between creative writing/roleplay exercises and what the user has shown they actually believe. I remember reading somewhere that to an LLM’s “world model” there isn’t much of a difference, such that it can get confused.

Idk there’s a lot to consider and you can go way deeper if the model is open source. Mainly, I just feel bad, because my stance towards people using LLMs is, more or less, “Caveat emptor.” Meaning people should know it’s gonna agree with them, and try to be a cheerleader for them, and if some misfortune befalls the user because they listened, that’s on them. But I don’t think that’s very tenable anymore, models need to be better 🤔

https://farcaster.xyz/stultulo/0xcceead92

great write up on evaluating sycophancy in LLMs

“With the internet and personal computing we can build networks with many interconnecting nodes that would be basically unstoppable.”

- Timothy C. May, Cypherpunk

Cypherpunks tattooed encryption algos b/c code was illegal

They printed banned algos in books & built tools to resist surveillance

B/c of them, encryption is legal

Roman built one, & now he faces prison

Support him if you can. The EF will match:

Cypherpunks tattooed encryption algos b/c code was illegal

They printed banned algos in books & built tools to resist surveillance

B/c of them, encryption is legal

Roman built one, & now he faces prison

Support him if you can. The EF will match:
https://juicebox.money/@defend-roman-storm

We’re co-hosting a hackathon this Thursday at Edge Esmeralda with Imbue AI!

It kicks off with talks from leading AI safety & d/acc thinkers and dives into building tools that protect, empower, and elevate humans in an AI-dominated world.

Read on + get involved 👇

d/acc discovery day cohosted by @octant talk from @vitalik.eth explaining what d/acc is!

so interesting hearing @vitalik.eth  speak on the purpose of d/acc! 

@octant

Slides from the TLDR GDrive -- all done in Cursor btw, hadn't written latex in a few yrs.

https://t.co/zSE0ZhHs60 https://t.co/MGJe3MoEvl

https://x.com/gakonst/status/1922058758974239194?s=46

anyone applying for a d/acc grant?

https://www.shiftgrants.org/

Excited to speak at the D/acc Human-Aligned Tech Workshop with @holonym! Join us to explore the future of human-first technology.

See you there!

privacy, security, and cryptography

https://holonym.id

Excited to speak at the D/acc Human-Aligned Tech Workshop with @holonym! Join us to explore the future of human-first technology.

See you there!

https://lu.ma/tfe9k2gf

How can we steer technological advancement in a defensive and decentralized approach?

At Edge City Austin, @vitalik.eth shared his updated vision for D/ACC, a framework for accelerating frontier technologies while maintaining safeguards.

Watch Vitalik's talk on YouTube: https://www.youtube.com/watch?v=04BacdXnNuQ

(We’re funding d/acc projects, more info in next post)

Free markets, strong governance, social safety nets, and innovation have driven prosperity in the physical world

My dream is that we can recreate it in the digital world where it doesn’t matter where you’re born anyone can prosper

popup-ama

这里需要先做个重要的纠正：d/acc 不是 de-acceleration，是 decentralized defensive acceleration 😛

这个是重要的，因为这个世界里确实有人支持deceleration，degrowth等等，但是我觉得这种方向是错误的，在和平的世界里会延误重要的医疗和基础设施的提升让更多人受伤，在现在的更危险的世界，如果不加速，就会被愿意加速的人吃掉

decentralized和defensive technology 需要跟其他技术竞争。如果剑快速进步，但是盾不进步，世界会变越来越危险。如果中心化技术快速进步，但是去中心化技术不进步，世界会越来越中心化。所以我们需要counterbalance这些趋势。区块链是这个故事的一部分，但是只是一部分，还有区块链之外的去中心化（比如，p2p网络），还有软件和硬件的安全（数字世界的“盾"），还有生物领域的很多事情，等等