Chris Lakin
Thank you. I started my blog because I was trying to make the post that anyone like me could read and self-cure—but then discovered that, even though readers liked recommendations, ~no one was actually getting better, and ~no one could pass my ITTs on my intended messages.
For as much as I believe in tracking results, I wasn’t tracking the results of my attempts at online guidance. So I checked, it was bad. I discovered readers couldn’t even self-diagnose (even after I’d spend a dozen hours trying to prevent that specific outcome), let-alone self-cure.
“Brain surgeon discovers patients not cured by reading blog posts about brain surgery.”
If anything, sharing scalable recommendations seemed to make things worse by implying there was asynchronous guidance that could lead to total success. I know many people who have been stuck for years because they keep trying to do the self-help thing, and the world is worse-off because for lacking their contributions. I myself lost a few percentage points of my life to this blindspot.
So what has worked for people? Working backwards, every big success story I’ve seen or heard involved ≥1 of:
Stressors or incentives disappearing: Mitigating a health issue, moving away, achieving financial independence, etc.
Supportive, loving relationships for years
Successful 1:1 work
Getting lucky with psychedelics on one of their first trips (?)
Which is to say: I only very rarely hear of big success stories involving people who did not have Maslow’s safety, people who relied entirely on self-guided approaches, or people who exclusively followed advice from people who themselves are destabilized.
Would you like more examples or context?
hm I think we should have an artist woman type choose
I really like the idea of branding this better, and I think we can do even better than what’s been proposed here. Let’s brainstorm in replies to this comment?
I want @Kaj_Sotala’s take or connections on this
A frontier lab researcher I worked with used to disagree with colleagues but say nothing. Two conversations later he was holding ground with people he looked up to. A year and a half later he still pushes back. I hunt bounties like this that improve AI alignment, DM me
I’ve seen this stem from need for validation (insecurity) instead of AIs themselves. A guy I know naturally stopped using ChatGPT 4h/day after his romantic anxiety disappeared. This makes the boundaries of the problem difficult. What are you going to do?— turn every AI into the best therapist? make every AI detect when the human is posting from insecurity and stop?
isn’t this true of humans too?
Have you seen Most “inner work” is not optimized for results?
Note: Sid is anonymous.
https://www.hyperstitionai.com
Aaron Silverbrook, today:
Hyperstition AI has shipped two open-source corpora (5,000 novels + 40,000 short stories, ~500M tokens), developed the generation pipelines, and collaborated with Geodesic Research on the experiments validating their alignment pretraining. https://alignmentpretraining.ai/
Our second corpus was to test Turntrout’s hypothesis for a filtered data set eliding all mention of AI and instead trying to convince the model that it was some kind of benevolent glass angel being https://turntrout.com/self-fulfilling-misalignment
Here’s both open-source corpora https://huggingface.co/datasets/jayterwahl/hyperstition
Thanks for sharing, surprised I haven’t seen more posts like this
the implementation took writers, copyeditors, web developers, backend developers, UX designers, a medical doctor whose patients were among our first users, and many more.
how much time do you think it would take to have made all of microcovid (including the research) today?
https://x.com/tomekkorbak/status/2038704753887379891
New OpenAI post: Can midtraining on docs about aligned AI bake in alignment priors for agents? We report an experiment where those priors are quickly washed away by RL and fail to generalize to agentic settings. But that cuts both ways: priors that AIs are misaligned fade too!
https://alignment.openai.com/how-far-does-alignment-midtraining-generalize/
(more in Richard’s thread)
in text form:
Richard Ngo @RichardMCNgo
The world’s most consequential Molochian dynamic is the race between AGI companies.
But several of the leading companies were significantly influenced by @slatestarcodex (e.g. as below)
So perhaps we should think of Scott as having *summoned* Moloch.
Quote
the blog post I think about most often: https://slatestarcodex.com/2014/07/30/meditations-on-moloch/
(other than a bunch of PG essays, which I somehow don’t count as blog posts)
It’s definitely not the most unlearning-ish algorithm there could be, but targeting unwanted responses directly is closer than not doing it
Fascinating haha. Currently, people reach out directly. I’m not exactly demand-constrained; also I’m concerned this then makes the posts look like an ad, and also it looks desperate to the people who matter. Curious what you think for integrating both of these constraints
I’m also clueless! I think the state of affairs is very unfortunate. I’ve outlined my standards for what people would do in an ideal world, but I don’t know of anyone else who meets them for internal bottlenecks imo. I’ve been pretty confused about what to do here, especially considering that big successes seem so rare