PauseAI•13mo ago

Reaching Out to Old Historical Contacts About AI Risk: A Guide

I decided I want to message my old historical contacts about AI risk, since it occurred to me that I have nonzero social capital with many of them and it's low-stakes if they aren't receptive. This appears to be a rarely-picked low-hanging fruit that we can help others pick. I myself am waiting to start this messaging until I gain more insight from others who help make the guide.

This is probably a half day of work for one person (or many hands make light work).

I think a guide requires these things:

An intro / explanation for what makes this a good untapped strategy for many people (perhaps including the reader)
A few examples of first-contact messages
A breakdown of what's important in a good first-contact message
Advice on how to guide the conversation if they respond
Wrapping that into a single short article

(If it's not fit for the website, it can be a Google Doc. In any case I believe it's very worth doing.)

full-greenOP•12/23/24, 4:53 AM

I just spent more than an hour just writing a short first-contact message. (I think many of you can do it better!)

Hi [Their Name], this is [My Name]. It's been a long while! I hope you're well. I'm reaching out all of a sudden because I've spent the last few years deeply following the progress in AI and AI safety, and I've only become more and more concerned about where things are headed. I've hesitated to speak frankly about this with people I know, but I've come to believe it's too important not to discuss.

What do you think about AI and where do you see it going? And have you heard anything about the field's safety issues?

hurt-tomato•12/23/24, 12:39 PM

I like this!

brilliant-lime•12/24/24, 6:04 PM

I like Reaching Out to Old Historical Contacts About AI Risk: A Guide, and what anthony did with a blog post.

I'm thinking I want to do more of a short convo starter. I think this might get people to engage more, but will cost me more time. Something like:

Hey [so and so], have you been following the AI situation? I'm getting pretty freaked out. What do you think of things?

I think telling people about PauseAI and X-Risk right off the bat might make some people disengage. With the more ambiguous start, I stand a better chance of seeing where people are at and guiding them towards concern and action, since I will get an impression of which direction they are coming from.

No matter how the conversation goes, I plan to mention PauseAI, X-Risk, and that I am focusing my career on them as much as possible, so even if people disengage they get the "people I have known think this is real" vibe, which will hopefully stay with them if it comes up in conversations for them.

I'm planning to do the spamming on Jan 04, and spend a few hours then responding to people who get back to me right away. I'm thinking of prepping a "text messaging ppl copy-pasta doc". I'll share it here when I do. I'll grab some content from PauseAI website, some of my own thoughts, and relevant links. I'll try to refine it as I spend time talking with people so it might be valuable to people doing this in the future. Do we already have anything like a copypasta doc?

brilliant-lime•12/24/24, 6:15 PM

I sometimes worry I might actually be socially competent. It's a scary thought :excuse_me_what: . I have weak evidence tho, people seem to universally identify that I am a weirdo, or "eccentric", or socially awkward, but I also seem to foster social harmony and be able to engage and communicate with diverse people... Being thought a fool puts people at ease

and it seems possible I'm really smart, but I'm definitely a fool

.

So, I'll try writing a draft of Reaching Out to Old Historical Contacts About AI Risk: A Guide when I work on my copy-pasta doc. I'm tentatively planning to do so on the 29th. Wish me a luck!

full-greenOP•12/24/24, 7:39 PM

Thank you so much, @trist! For encouraging, for facilitating, and for digging in and doing the work!

full-greenOP•12/24/24, 7:43 PM

A shorter conversation starter could be very good. It is more of an investment, but for those who can make that investment, we probably should.

When messaging people I haven't spoken to in multiple years, I probably need more of an acknowledgment of the gap and a tinge more justification for the message. But I should be able to cut down what I currently have so it isn't as intimidating.

full-greenOP•12/24/24, 7:51 PM

And of course, I'm very happy to help edit the guide once you have a draft!

brilliant-lime•12/30/24, 6:10 AM

I've started a document here:

brilliant-lime•12/30/24, 6:10 AM

https://docs.google.com/document/d/1w-j5W0by9ulqVIyfxfPhTRxlVJFGO7CtAsazhRF3aDU/edit?usp=sharing

Google Docs

Old Contact Outreach Draft 1

[ Epistemic status: First draft speculating on how this should be done. Anthony has some experience, but ideally, once people actually execute this plan, their experience will be fed back into the refinement of this document. I’m looking at you Tristan Trim and Nathan Metzger. ] What’s this? His...

brilliant-lime•12/30/24, 7:53 AM

Ok. I'm finished the first draft. I'll review it on or before Saturday when I start sending out messages.

brilliant-lime•12/30/24, 7:53 AM

If anyone want's to read and make comments here or there, please do

brilliant-lime•12/30/24, 8:44 AM

@anthonybailey.net I included a link to "Your personal copy of Doom" under "Examples -> Link Reference Ex 1"
Please let me know if you want me to modify it or remove it.

Also, if you have thoughts on the rest of the document I'd be grateful if you find time to share them

colossal-harlequin•1/3/25, 2:31 AM

No, completely fine to link. I only wish I could claim I know it works at all. In practice, I reach out individually to folk, cite that, add the obvious personal hooks, and suggest we meet up to chat. Around half my audience were colleagues current or ex from working at Amazon, and there were many lunch hour one-on-ones.

The pattern of success is that there is little pattern and mild success. Almost everyone gets off Liron's doom train somewhere before we reach the "action" destination. For any one contact, I am often very surprised by which stop they will turn out to choose. (Especially the ones after "yes I can see why we may well die.")

If cope within smart brains wasn't so deadly I think it might be the most wonderfully fascinating human thing I've ever encountered.

brilliant-lime•1/4/25, 9:50 PM

I'm starting my old contact outreach wish me luck

Bbrilliant-lime I'm starting my old contact outreach wish me luck 😎 🍀

full-greenOP•1/4/25, 10:10 PM

The final thing my todo list for today is giving feedback / first-round edits on the guide document. Sorry I've been slow -- I overcommitted myself at first. This month I'm gonna cook!

brilliant-lime•1/5/25, 1:31 AM

Results so far:

This is way more work than I thought it would be. Predictably, but I didn't predict. Messaging and talking with people takes time.
I really like the framing " I want your help growing a social movement I think is important ". It's a clear call to action which necessitates "understanding why PauseAI is important" as a step to support what I'm actually asking.

brilliant-lime•1/5/25, 1:33 AM

I'll keep working for another couple hours, but probably won't get through the contact priority list I put together. Might work on it some more tomorrow.

Bbrilliant-lime Results so far: - This is way more work than I thought it would be. Predictably,...

colossal-harlequin•1/5/25, 3:24 AM

I really like the framing " I want your help growing a social movement I think is important"

I second that, and note I find it also dodges some of the "la la la cope" issues. Some folk of ability engage with the abstract puzzle of growing an awareness/movement without having to push through the emotional cost of accepting the risk as real.

brilliant-lime•1/5/25, 3:38 AM

Lol, yup. Did you read:

This is a game. The goal is to get as many people to join the game as possible. If we win the game, the prize is a better outcome for humanity, and therefore each of us personally. If we lose the game, we may all die. This is not a religion. We are not blindly, deafly, dogmatic. If you become convinced we are misunderstanding the situation, it is important that we understand, and stop. But this conversation is older than you realize, and it is likely we are not misunderstanding the situation. Please understand.

This is a game. The goal is to get as many people to join the game as possible. If we win the game, the prize is a better outcome for humanity, and therefore each of us personally. If we lose the game, we may all die. This is not a religion. We are not blindly, deafly, dogmatic. If you become convinced we are misunderstanding the situation, it is important that we understand, and stop. But this conversation is older than you realize, and it is likely we are not misunderstanding the situation. Please understand.

from the document I wrote? I think it's kinda silly, but also kinda on point.

full-greenOP•1/5/25, 7:32 AM

@trist I finally read the guide and added a lot of suggested edits and comments. First thought: Wow, you are amazing. Thank you for putting this together! It's so good as a first draft.

full-greenOP•1/5/25, 7:33 AM

Keep that positivity in mind, since I did make a lot of suggested changes.

This is a game. The goal is to get as many people to join the game as possible. If we win the game, the prize is a better outcome for humanity, and therefore each of us personally. If we lose the game, we may all die. This is not a religion. We are not blindly, deafly, dogmatic. If you become convinced we are misunderstanding the situation, it is important that we understand, and stop. But this conversation is older than you realize, and it is likely we are not misunderstanding the situation. Please understand.

Reaching Out to Old Historical Contacts About AI Risk: A Guide

Similar Threads

Similar Threads

Similar Threads