Influencer outreach (e-mails)

[rich imp]

We won't change any opinions if we only talk with each other. We need to get out there, and find other people to convince that a Pause is the way to go.

One way to do this, is through influencers. This means opinion writers, video creators, journalists... People who made it their job to create and share content for others.

@gleaming hill has started a project like this earlier, and we worked on creating a list of Youtubers that can be mailed.

How to reach out

• Find their email (often quite easy, on their youtube channel / website)
• Know what they have created and what they write about, especially about AI stuff
• Come up with an idea for a video / content that fits their platform
• Give a sincere compliment about what they're doing
• Give your suggestion
• Suggest to help out
• You can mention that you're a volunteer at PauseAI, and that many people there would like to help out
• Give sources for your claims
• Write down a comment here below to share your progress!
• Suggest a timeslot to meet

I've reached out to Kurzgesagt (Philipp Dettmer) last week, have sent a reminder mail just now.

[gleaming hill]

I have no problem with people writing (decent, serious) letters, but maybe you should coordinate here so that your targets don't get spammed with a flood of letters - this might not be instrumental to our cause.

I intend to write a few letters, soon, and I will mention it here when I do, for that reason.

If anyone needs advice on how to approach someone or on what sources might be appropriate, feel free to reach out to me and I will try to help.

[rich imp]

@blissful flax We had a chat last week about reaching out to some people, too. Perhaps instead of making a list in a sheet file, we can just put the names in this thread. A sheet may be a little overkill for smaller numbers

[blissful flax]

OK sure. It would be good if we can have a quick call sometime about this. Let me know when you might be available.

[gleaming hill]

@blissful flax suggest a time and I will say if I can participate. Do we have other regulars that might want to join?

[prime rapids]

@desert bluff

[desert bluff]

Oh, thanks Greg!

@rich imp You got a message regarding Kurzgesagt. I now have a contact there and there are some promising developments that I for now don’t want to divulge to the whole server. Maybe next week.

[rich imp]

Promising! 😄

[nova nest]

By the way I used to work for a big campaign, and we subscribed to a database with all influencers, including their contact details, charity affiliations and interests. It was pricey but it did work successfully for attracting some big celebrities to our cause. The subscription fee has probably increased now because this experience was years ago.

[rich imp]

Sounds interesting! What's this called?

[nova nest]

I believe it's now https://www.influencerintelligence.com

[silver prism]

just found this. did you contact someone @gleaming hill ?

[desert bluff]

Just to update on Kurzgesagt: I met the founder at EAGx Berlin and messaged with him. On stage he described how hard it is to condense everything in 10 min. (and he's right ... I am trying this for weeks now). He seemed still optimistic, but in private he admitted that they are not working on the topic in the near future. If he picks it up again, he will contact me.

[barren pier]

Perhaps a lite project management tool could be helpful to track actions, due dates, follow ups etc

[rich imp]

Yes, true. Can you create a chart in Google sheets for example?

[rich imp]

@hallow depot mentioned Philip de Franco as a target. He's posting a lot of news, including some AI stuff calling for guardrails

[hallow depot]

specifically, saying he commends anyone putting in guardrails. he generally talks about deepfakes a lot, both because he's a creator himself, who's been deepfaked and because he's been covering the SAG-AFTRA strike

He also talked about the Biden Executive Order

[frank trail]

random idea, what is the opinion of Randall Munroe (XKCD) on x-risk?

[rich imp]

Great idea! He actually wrote about robot apocalypse earlier: https://what-if.xkcd.com/5/

TLDR: He thinks robots will fail

[hallow depot]

The superintelligence will be dumb

[silver prism]

I've started making a list analyzing different creators to see what seems like the better choices but really the biggest risks that I see happens if the creator is controversial, partisan or close minded.

So for example we could contact other EA creators like 80k hs podcast, AI X-risk Research Podcast, the FLI podcast, the inside view... In case somebody would be willing to go.

@rich imp I know you've been in some podcasts, have you already contact those?

(btw i didn't mention Rob Miles because he have already been in contact with some people and I understand he's not interested)

[hallow depot]

to be fair, i think this was quite some time before chatgpt

before that, i agreed with the comic- knew a lot less then though

btw, with the news forum and videos, is a chance for pauseai to become a popular channel

my time for that isn't going to be the best, since i have other responsibilities as well, but if someone has more time, they could definitely do it!

[rich imp]

Haven't reached out to them yet!

[silver prism]

Should I do it? I don't feel super confident when it comes to communicating "profesionally" but I can do it anyways

[rich imp]

Please do 🙂

[silver prism]

I'm writing an e-mail to 80k hs and will send to the others that I mention later. dont send it yourselves

[woven yarrow]

@silver prism is the suggestion to 80kh basically "you've tended to emphasize technical alignment and governance to give more time for that, compared to what it might take to enact a longer term pause if necessary - can we discuss the latter some?"

[silver prism]

I don't know what their opinions about a pause are, the mail that I sent was about making a podcast episode about it with Joep and/or someone else, I even mentioned that they could be someone who is against it to have more of a discussion (in case they are unsure about it)

[woven yarrow]

(Thanks, sounds cool to me.)

[silver prism]

Rob Wiblin answered! He said he'll seriously consider doing an episode on the 'Pause AI' topic after febraury :DD

[rich imp]

Awesome!

[silver prism]

This is still inside the EA bubble but I think it can open paths to bigger podcast and stuff

[hallow depot]

Let's draft a template which we can mass send

Hi, I'd love to talk with you about PauseAI. We're working to makre sure that AI doesn't kill humanity. Promise we're not nutjobs. 

Best ___

[spring slate]

"doesnt pose serious threats to humanity"

might be more diplomatic

[cloud moon]

Possibly off-the-wall proposal: reach out to popular authors, actors, directors, musicians, etc.? Obviously these people get a ton of email from all sorts cranks, and they have no more credibility or insight than anyone else on the x-risk issue. However, they do have an obvious reason to dislike and fear generative AI, which poses a double existential threat to them. There are more than a few
famous creative figures who have already expressed anti-AI sentiments. It's a long shot, but it would be great if some beloved big-name celebrity could be persuaded to use his or her platform to spread the message that AI is profoundly socially damaging and potentially catastrophically dangerous in ways that we simply are not equipped to effectively mitigate.

[silver prism]

yeah, that sounds like a great idea

we should probably expand a little bit more https://pauseai.info/risks, specially the part about job loss and introduce those people through it

[cloud moon]

My thinking about all of this is that, if we continue on the current path, a good outcome is very unlikely. The bad outcome may be a dystopia (extreme wealth inequality, ubiquitous surveillance, algorithmically institutionalized prejudice, everyday human interactions significantly replaced with chatbot interactions, etc.) or it may be the apocalypse (misaligned AI kills us all), but either way we don't want it.

I think an exclusive focus on x-risk doesn't necessarily pull people in.

[silver prism]

I think the only reason why I haven't been saying "let's mass e-mail everyone" whether influencers, politicians, or funding. It's because first impressions are important and we can always better the website or the server (to retain more people)

Also there's a point that if some % start to respond maybe you lack the time to respond to everyone. but we're far from that

[cloud moon]

Re: the website—maybe something close to the current overview page, but each of the points links to an expanded version?

And move jobs/wealth inequality closer to the top? That touches on a near-universal fear.

[silver prism]

yeah, that sounds good. if you want to give more feedback #1138493909778051234 or you can do it directly in the GitHub

[cloud moon]

Awesome, thanks!

[silver prism]

Which famous creative figures have expressed anti-AI sentiments? I don't follow artists and people like that much

[hallow depot]

Bill Burr

[cloud moon]

George RR Martin and John Grisham are suing OpenAI for using their novels to train the algorithm

I'd take that as an indicator of anti-AI sentiment

A few years ago, Miyazaki Hayao famously flipped out about an AI-powered video game

"I feel strongly that this is an insult to life"

[silver prism]

I think that if you wanna try it, you guys that know more about them you should definitely go for it.

[silver prism]

I made a list of people that I think we could contact but also have big following and I think there are not a lot of chances with them. There's a couple exceptions tho.

The person I think the most of all by faaar is Destiny, a streamer/ youtuber who I think has big pros BUT also a big con and is that he can be really controversial and political, although I think the people who hate them are from the extremes. He can become a douchebag but I'm pretty sure just with people that treat him bad on the first place. I don't know if you know him, he streamed the Manifold Conference and """debated""" with Eliezer.
The pros are:

1. He seems quite accessible to have a talk if he's interested on the topic (that could be the case). He's really open / philosophy minded so given I think we are right about our beliefs I believe he will agree with them too.
2. He has a big fanbase that is REALLY strong. They really care about his opinions and have gone canvassing for political reasons and could go protest I believe. Also his community is really active on Discord and Reddit, so they could be really active here too and participate in other projects.

I'm not sure why my intuition is so strong about being really good to talk with him. I really don't feel like if he were to support a Pause and talk about it he would politicize it. I feel like polarizations are less based on ad hominems and more in narratives.

I mean, Eliezer went to a left wing guy (David Pakman), to right wing guy, and nothing happened because of that. I guess that there is a world where he could become pretty vocal about it and that could be bad but we could also ask him to not do that idk.

.
Sam Harris is pretty hated and vocal of AI risks and I don't think drove hate to it ? Destiny was invited to the Manifold thing and they weren't afraid to politicize prediction markets.

.
David Pakman who I also mentioned and also is a friend of Destiny covers news and stuff and seems to believe in AIXR so it would be good to try to make him cover a protest or interview someone too.
Again, the con is that he has a political lean, but I feel is not that important.

[hallow depot]

I think we can reach out to any

Hi Philip, long time fan (I remember the glove 😄)
I'd love to talk with you about PauseAI. We're working to make sure that AI doesn't kill humanity. Promise we're not nutjobs.

Best, Kabir Kumar

just sent to philip defranco

Hi Louis,
Thank you for everything you're doing for Right to Repair- both with lobbying and the videos. Your videos about the lobbying and Charlie Brown were what informed me about lobbyists, how they talk and some of the things to watch out for.
Do you know about PauseAI? They're trying to make sure AI doesn't kill everyone. They're not not nutjobs, I've talked to a lot.

Best, Kabir Kumar

just sent to Louis Rossman

gtg, will do more later

Let me know if these are bad

[silver prism]

imo a little bit too short and changing topics so fast can be seen like the first part it's just to manipulate them to feel sympathy or smth?

I really like the second one, and also I think that guy could be interested, I didn't think about him, prob because I don't know him that much

but seems like a good shot

[hallow depot]

second one is prob better because I watched a lot of his videos and they did teach me important things

[silver prism]

Now that I think about it I'll add that they can check our website to learn more about PauseAI and the risks,

[silver prism]

Most of the people I think about do podcasts type of content, but I don't feel comfortable at all participating on that. So any good communicator that it's interested on promoting a pause is super valuable. Like Joep and Holly. I wonder what's Connor Leahy opinion on PauseAI, given that he supports StopAGI I think.

I will of course ask before sending someone Joep and Holly info if they would be interested.

[silver prism]

3 out of 3! EA podcasts have shown interest. Next week I'll probably send outside the bubble.

(the other 2 being michael/ the inside view and FLI)

[rich imp]

Amazing work @silver prism !

[rich imp]

I'd love to!

[woven yarrow]

Joining dots, going to note the kind of big bet touched on in #quick-chat-💬 message - popular TV shows.

[rich imp]

@zealous mulch Arguments for acting on short timelines are IMO very easy to make:

• It is hard to predict and the stakes are high, we need to err on the side of caution. If there is even a 1% chance that AGI is achieved in the next 12 months, that already would constitute an unacceptably high risk.
• Rate of progress keeps being underestimated. A 2022 survey asked AI scientists to estimate when the first AI written book would be a NYT bestseller. The estimate was 2039. The same survey was held one year later, the estimate was lowered to 2030. But the real kicker is that in december 2023, a Chine professor already won an award with an AI written novel.
• We suck at exponentials. Humans always undereestimate exponentials. This is a proven scientific fact.
• There are so many factors that are accelerating. We're not just spending more money on training runs (factor 1), but also designing new chips optimized for ao (factor 2), chips are getting smaller and faster in general (factor 3), we discover novel algorithms / architectures for training (factor 4), new runtime innovations (factor 5)... I'm probably missing more. The point is: perhaps even individually, one of these could get us beyond some dangerous threshold (e.g. being better than an AI researcher at AI research) in a relatively short time. Combined, even more so.

[exotic forum]

PauseAI banner drop at a Taylor Swift concert

[rich imp]

Reaching out to taylor could work, if she's all like "whoah AI sucks amirite" I suppose that could help out a lot. She is the most popular person alive right now.

[silver prism]

We should add those risks to the page for that (deepfakes) that I PRed

[hallow depot]

Jon Stewart seems very good

Thoughts on Joe Rogan?

Tristan Harris has already been on his podcast

[silver prism]

Yeah, I feel like the person who could go there would be Max Tegmark? Who has gone to Lex advocating for a pause (althought maybe it was the one just for 6 months). It would be really good to have Max supporting PauseAI in case he goes to more podcasts

[hallow depot]

Pretty confident he'd like the idea

[rich imp]

Something like "AI risk mitigation"

[hallow depot]

Fantastic!!

[silver prism]

Also there is this dangerous idea: contact a swiftie group ?

[silver prism]

how to best describe PauseAI in these e-mails? a community? a movement?

a small movement?

[silver prism]

now that I added the Deepfakes stuff to the page, this night I'll send messages to Gibi ASMR (made a video against them), David Pakman, Taylor (lol) and Destiny

[rich imp]

All of these descriptions are good. Activist group, movement, community... I often say "community of volunteers"

[hallow depot]

I think we should also reach out to more 'right-wing' groups

so it doesn't become partisan

maybe right wing/conservative journalists

[silver prism]

I don't know who tho

just the guy who Eliezer talk with, Dan Crenshaw

[hallow depot]

Sean Hannity

Also, the guy who was speaking to the white house about AI being an existential risk, who first got laughed at, then months later, got a serious face

[frank trail]

@hallow depot his name is Peter Doocy and he is a fox news correspondent

[hallow depot]

Sweet, thank you, I'll reach out today

Not a problem

Thank you, though

I'll probably frame pause ai as something like building our defenses to protect our country's interests

Choosing the right time, not being at the whim of tech companies

Etc

Will watch some interviews, shows, read articles first

That's kinda framing myself as an opposition

So not the approach I'd take

Ideas are still great!

Talked with some lawyers recently

Wild how much they didnt know about even the absolute basic baby stuff about ml, let alone ai

But had wildly confident opinions

We're in a bubble, inside of a bubble, inside of another bubble

[hallow depot]

Was talking to lawyers generally, for the Law-a-Thon

[silver prism]

So, I noticed that writing e-mails take more time and text than I thought when it is to people who you don't know what they know but probably not much.

.
Also I've been thinking that the best thing to do right now maybe would be to #1196467528172584960. Because if, for example, an important figure that we're trying to reach out give us a platform, that could cause a influx from their community which I don't know if we would know how to handle

And having like a staff and leadership from EA minded people, who for example wouldn't want to risk politicizing stuff too much, would be probably good. I mean we already have that, but it would be cool to have more people like that

[warm thorn]

Taylor Swift team's contact info for fans: https://www.thefandom.site/taylor-swift/

Hope this helps, team!

[silver prism]

Tysm, I actually couldn't find this before

LMAO

plot twist

[hallow depot]

we need to reach out to Lindsay Graham

Senator Durbin

[silver prism]

Influencer outreach (e-mails)

[silver prism]

0/5 btw. obviously bigger figures and who haven't expressed concerns about AIXR are more difficult.

I still think is pretty good to try with most famous people, given one good communicator could lead to a lot more

another tactic to get to them could be to get through other contacts of them instead of their e-mails

Honestly tho, I've been thinking that the most important group to convince is right wing people. Yeah, some of them already agree with them to an extent, but we need to get them on our side

I don't think left wing AI ethicist people will get against us, they could not help us but i don't think they will get against our proposals

[oak hemlock]

I decided to take a more direct approach and try to setup a meeting with my congresswoman, but I have no idea if she'll bite. If you read the attached letter you'll see that I offered to give her a formal in-person briefing on AI Risk, either in her DC offices or PA. Her committee assignments aren't a a good match for the topic, but AI is such a hot topic she may still be interested. Fingers crossed.

[rich imp]

I've sent an email to Adam McKay (guy who made Don't Look Up and The Big Short). Not asking him for anything, just want to let him know that he's awesome and that he was right about humans.

[frank trail]

haha, what did he say about humans?

[rich imp]

Well, the whole Don't Look Up movie paints a very negative image of humanity. Incapable of acting. I thought the movie was very unrealistic when I saw it. Turns out I misjudged most humans, and the movie is actually quite accurate.

[frank trail]

hahaha, well we will see about that, the movie we are in is not over yet 😋

[rich imp]

True 🙂

Nice! He even did a podcast with Max Tegmark: https://youtu.be/zeRe2m3ZbT0?si=J_eLMbTY8qRpzAIS&t=824

[woven yarrow]

Noting a tiny tease from Jon Stewart on last night's Daily Show that this year "maybe we'll talk about China, maybe we'll talk about AI".

I think we didn't manage to locate any random connection into the research team but rattling the cage here once more just in case.

[oak hemlock]

Hi! My state representative just did a phone blast inviting Pennsylvania Distric 163 district residents, like myself, to attend a educational funding meeting on the 20th. I'll attend and try to introduce myself, and then try to setup a quick PauseAI meeting on a later date. She's a member of the state's Labor & Industry committee, which sounds ideal for an "AI's Taking Our Jobs" sort of approach. Unfortunately, that's my least schooled area--I'm much much better at existential risk--so I could use all the help I can get. Any suggestions / resources / help would be greatly appreciated. Thanks...

[warm thorn]

You may find the video of our nonprofit StakeOut.AI's webinar for Hollywood actors, titled "How to prevent AI from stealing your job (not an official SAG-AFTRA event)" a useful resource.

Hope this helps, @oak hemlock !

[silver prism]

Btw David Pakman does callins in his discord server and people have gone to talk about risks of AI, somebody could go but idk how to frame it to not look as a self-promotion thing ?

And then some of those calls end up in his YT channel

[hallow depot]

Signs of mrbeast looking down on safety in general (not ai- general workplace)
So if he's approached, safety won't be the angle that works

See: https://m.youtube.com/watch?v=EWAJl_GgwXU

Also, he's interacting with altman quite a bit on twitter

[warm thorn]

Oops, forgot to link, @oak hemlock!
https://www.youtube.com/watch?v=gr_BlCs_04Y

[rich imp]

but maybe he's a little concerned about other stuff

[hallow depot]

Yup- not saying he can't be reached, just theorizing that safety isn't the word/angle

[silver prism]

Successful emails

[frank trail]

oh wow @rich imp may go on the 80k podcast?? 🤩

[silver prism]

Holly may

[rich imp]

Also makes more sense, because she's well-known and respected in the EA scene!

[prime rapids]

Oh wow, would be great to see! @mellow yoke can you plug my funding circle on there if possible? 🙂 (could maybe relate it to this point Kat Woods made recently about existing big AI Safety funders having conflicting interests: https://twitter.com/Kat__Woods/status/1759019889413820607?s=19 )

Kat Woods ⏸️ (@Kat__Woods) on X

Many big AI safety funders can’t donate to AGI pause initiatives, so it’s a great spot where smaller donors can have a lot of counterfactual impact.

This is because if you fund capabilities orgs or advocacy groups that are the “good cops” towards AI labs, you can’t also be…

Twitter · Feb 18, 2024, 12:59 AM

[woven yarrow]

Randomly noting #1209192184792748101 message since she's popular and beginning to engage more.

[mellow yoke]

Yes, absolutely!

If that happens, ofc

[frank trail]

I don't know who that person is but should we contact them?

[silver prism]

Yeah

[frank trail]

Has anyone approached Kurzgesagt already?

They just posted a video on nuclear war supported by OpenPhilantropy that got 500k views in a few hours. Imagine the impact if they decided to do the same for AI X-risk (I wonder if OpenPhilantropy is also working on this) https://www.youtube.com/watch?v=LrIRuqr_Ozg

[silver prism]

yes

[silver prism]

.

I wish they would just make multiple videos

[silver prism]

#media-📺 message will write to captainsparklez this week, but if someone can do it before be my guess

[humble breach]

https://youtu.be/5DDX_ACqneU

[humble breach]

He's done a bunch of videos about very thoughtful, relatively concerned takes on AI. He's probably not the kind to do a sponsored video, but encouraging more like it or asking for a video link would be great.

Sam from Wendover also seems like the kind of person who would do a video on the risks of AI.

[silver prism]

We don't sponsor anything

[oak hemlock]

I had a great first meeting with with US State Representative Heather Boyd and got to speak with her for almost 10 minutes. It was little more than a bit of grip-and-grin, but the point to it all is that her Legislative Director will be hooking me up with an appointment (for next week!) And, BTW, Heather was the one who first mentioned AI; not me.

My two-part goal is to introduce Heather and her team to Pause AI, followed by my being introduced to some of the other politician's in the state. It turns out that Heather is quite close to Mary Gay Scanlon (my congresswoman; whom she worked for as Senior Advisor), plus she's definitely worried about the effects of AI on the job market. She was also interested in reading Darren McKee's "Uncontrollable" (which I gave to her!)

I could really use some help in solidifying the messaging for that meeting, so please send your ideas and/or reach out to help me brainstorm.

[humble breach]

My bad, I was thinking about a video about a specific organization and not a concept.

[glacial horizon]

I was thinking that it might be beneficial to contact a few artists that are quite well known in the anti ai art community, such as Deb JJ Lee, and Jonas Jodicke. There is also the Human Artistry Campaign that advocates for generative AI being developed more responsibly to protect creators' interests. I think that might be a beneficial connection?

[silver prism]

Yeah, that sounds good. But be clear that a pause on frontier models wouldnt affect today's AIs

[glacial horizon]

Yeah, I'm not so sure about the specific overlaps but I'm sure they'd appreciate the sentiment

[silver prism]

Today DaganOnAI uploaded a segment of his last video https://www.youtube.com/watch?v=u_dqC3T2kCw. it would be good to contact him in case he wants to collaborate either making video(s) about PAI/ a international treaty to put a cap on compute or at least mentioning it as a strategy

[woven yarrow]

(The longer Moloch piece on the same channel links the PauseAI action page among sources at the end of the final section on co-ordination.)

[silver prism]

What type of influencer are you thinking off? Is getting paid for voicecalls common in some circles?

[rich imp]

Of course!

[glacial horizon]

I've reaching out to a few artists but now i dont really know what to do..

can someone advise me on where to go from here...

[frank trail]

Is that their last reply to you?

[glacial horizon]

Yeah

[hallow depot]

is it ok to say which companies it'd be from?

[glacial horizon]

its just one artist

[humble breach]

https://youtu.be/JrcbH0ge2WE?si=q6fUk0Q_A47wsnXB

[oak hemlock]

QQ: I want to raffle off five existential-risk books at John Sherman's forthcoming "Thrill Us, Then Kill Us" talk (https://www.meetup.com/philly-net/events/298710679/) but I'm unsure which book(s) would be best to offer. I'm tempted to give away Dareen McKee's "Uncontrollable," but it'd perhaps be a better idea to offer a mix of titles. ANY IDEAS? Also, is there a a list somewhere of all such titles?

[jolly jay]

There’s https://pauseai.info/learn#books with the latest additions being mckee and yampolskiy

[silver prism]

I just commented the one from yampolskiy because someone in the server put it was kind of academical and haven't heard other opinions

but if someone else reads it and recommend it I will uncomment it

[jolly jay]

Didn’t read it yet either, I’ve just noticed that everyone was talking about those two books

[hallow depot]

personally, I'm skeptical of a lot of dr yampolsky's arguments

a lot of them seem kind of motte bailey to me

[woven yarrow]

Where in general the motte is "mathematically proved impossibility theorems" and the bailey is "which translate simply to the real world"?

Have you already described this in more detail anywhere @hallow depot?

[woven yarrow]

Separately: I'm sure others can do better than me, but prodding Last Week Tonight: https://fxtwitter.com/anthonybailey/status/1765153784798323177

[zealous moat]

Do we have a CRM for this?

[rich imp]

No, not yet. #1213591270992379935

[silver prism]

now you have to send two more e-mails then? 😅

[rich imp]

https://youtu.be/JrcbH0ge2WE?si=2ZJKWF9O_qlLKvmO

Kyle hill on genAI

[humble breach]

Yeah, I already emailed him about this video

[humble breach]

I mean, I actually emailed his agent because I'm off social media. The best option is probably his patreon or discord.

[frank trail]

Hey everyone, I would like to figure out how effective this project has been so far. Anyone knows if one of the influencers that were contacted has posted something about AI risk as a result of this contact?

[frank trail]

Im about to send an email to "Joueur du Grenier", famous french youtuber that I've been following since I was a kid 😍

Salut JDG!

Je suis un énorme fan de ce que tu fais depuis tes toutes premières vidéos (on a pas un énorme écart d'âge)!

Je te contacte car j'ai bossé dans l'IA pendant 10 ans et je fais partie des chercheurs qui sonnent l'alarme sur les développements rapides actuels. J'imagine que tu as dû voir SORA le tout nouveau modèle qui crée des vidéos réalistes? Je me demandais ce que tu pensais de ça, si ça t'inquiètes, et si c'était un sujet que tu serais intéressé d'adresser dans une de tes vidéos? Si jamais c'est le cas et que tu veux obtenir de l'information sur ce qu'il se passe dans ce domaine, je serais absolument ravi d'avoir un échange avec toi, ou de te mettre en contact avec des experts!

C'est un peu une bouteille à la mer vu la quantité d'emails que tu dois recevoir, mais on sait jamais 🙂

Salutations,
Maxime Fournes

PS: j'arrive pas a croire que j'envoie un mail au JDG :)"

[frank trail]

To Alex O'Connor (Cosmic Skeptic)

Hey Alex!

I am a massive fan of yours. Among other things, I became vegan in great parts thanks to one of your videos. I thank you for this.

I am one of the many experts in AI who are sounding the alarm about the massive rate of progress and the risky implications for humanity. I assume that you have already thought about this problem much, and are aware of all the recent developments, but I thought I would contact you just in case. I could offer information and put you in touch with experts in this domain if you are interested in learning more or addressing this topic in your videos.

Love and kindness
Maxime Fournes

[silver prism]

I think there haven't been responses outside of the pinned EA podcasts. Outside of that with regular influencers of different sizes I'm 0/6. I have two e-mails written and a lot that I'm thinking to write this week but I wanted to have the /feasibility page published before sending them.

[frank trail]

OK I have been reading through this project thread and collecting all the information. I want to do a little debrief of this project to see what we can learn already and how we can improve the process, as I suspect this project is extremely high impact. Convincing just one influencer could mean a massive instant growth of Pause AI and AI safety. I will ask people who have sent emails to check the information I have collected and give updates

@rich imp

• Kurzgesagt: In touch, any update?
• Adam McKay: any update?
• Any other?

@hallow depot

• Philip de Franco: no response?
• Louis Rossman: no response?
• Peter Doocy: no response?
• Any other?

@silver prism

• 80k: success, any update?
• AI x-risk research podcast: success, any update?
• FLI podcast: success, any update?
• The inside view: no response?
• Gibi ASMR: no response?
• David Packman: no response?
• Taylor Swift: no response?
• Destiny: no response?
• Captainsparklez: no response?
• Any other?

@zealous mulch

• Secular Talk: reached via discord. Update?
• Owen Jones: no response?
• George Monbiot: no response?
• Gary Stevenson: no response?
• Double Down: no response?
• Game devs: which ones? update?
• Any others?

@glacial horizon - A few artists: which ones? Had at least one response. Update?

[hallow depot]

No, no response from defranco or rossman. Doocy, I havent reached out to

[frank trail]

@humble breach

• Kyle Hill: no response?
• any other?

[humble breach]

No, no response. He has a patreon discord, but it costs $10/m or something. I haven't done any others yet, since I've been writing to representatives and trying to plan a protest.

[glacial horizon]

Not particularly, sorry

[frank trail]

No worries at all, it's all for the learning. How many people did you reach out to in total? You got only one response?

[glacial horizon]

I think 4

Just DMs on insta

[rich imp]

Kurzgesagt and Adam McKay both never replied unfortunately

Oh wait - the Adam mail was never successfully delivered... I should try again 🙂

Thanks @frank trail for checking in this is useful

[frank trail]

@desert bluff any update on your contacts with Kurzgesagdt?

[silver prism]

80khs updates: they maybe contacted Holly or will in the future, idk
AIXR: I didn't end up contacting them given they are about research
FLI: episode done
Inside View: episode done
The rest no response but I didn't contacted Captain Sparklez, but BestGuyEver. AlvaMajo & DaganOnAI.
So I guess 0/7

[desert bluff]

Yeah … they unfortunately have given up on it for now. They don’t see a way to fit everything into a 12-15 min. video. And I feel them: I can’t either 😅

[rich imp]

Honestly, what a lame excuse not to make a video about it. We're talking about every single life on earth.

[silver prism]

I'm sure they're not that doomers
But anyways, I don't understand why they don't just do a longer video or split it in various parts.

[hallow depot]

yup

maybe we make a script for them?

although, aisafety.info has probably got stuff we can use

[glass jewel]

"so that your targets don't get spammed with a flood of letters - this might not be instrumental to our cause."

I think it's the opposite effect. If you receive one email from somebody suggesting you do an episode on a topic, you might be persuaded. If you receive 10, especially if they seem organic (which in this case, they would be!), you're going to be like "wow! There's a lot of demand for this. Maybe I should do an episode on it"

[hallow depot]

Yup. E.g. if I get 20 different requests for a feature on the site I go 'holy shit, people really want this, I should do this straight away!'

[crimson zenith]

If I can help with that writing emails to show that there is demand I can do it

[silver prism]

I would say: If you know people on the lists, e-mail them

And you like them I guess

[silver prism]

reminder

I think you can ask him about his opinion on the AI commissioned report and he is open to talk to all type of AI worries, even talked with Eliezer and Sam Harris about it

maybe he didn't read about the report

Although politics is getting big now and he is really partisan so maybe is better to forget what I've said 😅

[hallow depot]

i'm down - link to the server?

[silver prism]

https://discord.com/invite/DZcYXhy

[silver prism]

Multiple e-mails to the same content creator

Okay, so if the multiple emails theory of impact is real, how should we coordinate to e-mail someone on particular? After deciding someone, should we send all mails in the course of a week, two?

[jolly jay]

A steady flow would probably be better and seem more organic than a single burst in that case 😅

[silver prism]

yeah, probably

any good suggestions? I worry we don't watch the same youtubers or podcasts

I mean, we don't need to all watch them, if we think they are kinda worried by AI I think that's good enough to contact them

In the non so serious side, I wrote an e-mail to jREG, who is known for making comedic videos but has also done honest sincere ones and lately has talked about the singularity

[slow hull]

YES THE COLLAB WE NEED

[hallow depot]

Philip DeFranco?

he seems to pay a lot of attention to what viewers are saying

What are some conservative youtubers/influencers?

Candace Owens

Dan Bongino

[silver prism]

I guess that there are three big levels of collaborations that I'm seeing possible:

1. They do content on some AI risk
2. They mention a pause/ PauseAI / they endorse it
3. They have someone from PAI to talk about AI risks and PAI

Those people that you mention have already talk about AI risks, and don't have people on.

So I guess you would want them to talk about a pause (point 2) or AIXR in particular (point 1).
.

The only perspective from where they could cover 2 given that they talk about news is sending them the commissioned report.

Which sounds like a great idea.

I was thinking that I wouldn't like how those people would cover PAI or AIXR but they could cover those news nicely

Pretty sad that a week have already past since the news, I think we should be prepared next time to send to these type of alternative media outlets

I will try anyways

[frank trail]

Was taking a few days break, but planning to do a bit more strategising on this topic. One thing I am considering and would like to discuss is whether we should mention PauseAI in the first email. I think that if multiple people mention pauseai when emailing someone it looks like a coordinated campaign. If we dont mention it immediately it looks more organic and could have more impact

[frank trail]

https://youtu.be/2hlD7dWp09M?si=9Uz69gZSugLmvLb2 we need to contact this guy. Will get on it

[silver prism]

I wish you luck. It seems the kind of channel that try to look unbias, smart and unemotional just making jokes here and there and talking in a monotone voice without giving opinions that could look emotional.

[frank trail]

I have followed for some time and I have noticed his point of view evolving on ai, roughly from "this is an obvious marketing scam" to "this is concerning"

[frank trail]

Email sent.

Hello Fireship!

I have been following your Youtube channel for a few years now, and I really enjoy your educational content. I have noticed that you have been producing more videos on the topic of AI lately, which prompted me to write to you.

I am one of the many experts in AI who are sounding the alarm about the massive rate of progress and the risky implications for humanity. I used to be a researcher in machine learning and deep learning, most recently leading a data science team at Two Sigma, but I have become increasingly concerned about the rapid progress and decided to leave the field. I am currently working in AI safety and doing voluntary work for Pause AI, a grassroots movement calling for an indefinite pause in AI research.

If you are interested I would be happy to provide detailed information and put you in touch with other experts who share these concerns. I believe that you are in a unique position to raise awareness about what is happening and the dire implications for all of us.
Please let me know if you would like to discuss this further.

Kind regards,
Maxime Fournes

[slow hull]

ILL SEND HIM ONE TOO AND TELL HIM IM WITH MY BUDDY MAXIME!!:)

I am gonna reach out to the good people at “some more news” and “the majority report”

[hallow depot]

Awesome!!

[silver prism]

The one and only Jordan Peterson has expressed worry about lots of risks, including lose of control I think. [email protected]

[oak hemlock]

I wrote a letter to Sam Harris:

I’d normally get into it slowly, introduce my subject mincingly—we’re strangers, after all!—offer caveats, soften the blow and try to be “balanced,” but the time for half measures is rapidly drawing to a close. “ASI is on a path to murder your children!” And me, and you, and ...well, everyone; every single person on the planet, probably even the gardenias. And soon!

I’m not quite sure what your thoughts on the matter might be, but I’ve been encouraged by your previous words on the subject, beginning with your famous TED talk.

Anyway, the point to it all is that words matter. I like your words; that’s why I’m writing.

I need that most rare of rare rareties: a 25-word elevator-pitch that’s perfectly crafted to wriggle it’s way into every hearer’s limbic system; especially that of government officials. The object: to make the urgent case for pausing frontier-grade AI until humanity has a better grasp on AI-safety and thereby a more likely chance of not being murdered.

As to why “I,” Louis Berman, need such a thing, I’m embarked upon a PauseAI project to give a “Gov’t Action Kit” to each and every US politician, to get the message out and advocate for real change. The kit will help the conversation, but I need a so-called elevator-pitch to even get in the door. And there are a lot of doors—600 federal politicians plus 200 media outlets, for a start—so it better open a bunch of doors.

As to my actually obtaining your help, I have no idea if you’d be so inclined (let alone that you could even be convinced that the danger is real), but if you find-yourself willing to help, or to even learn more, please email me.

In either case thank you for your attention. One can only imagine how many unsolicited emails you must receive each and every day…

[frank trail]

Awesome! Was about to also reach out to him, I will wait one week

[silver prism]

if someone would be down to go to a show / interview about their problems with AI / worries, you should try to contact this channel https://www.youtube.com/@syrmor

[oak hemlock]

Here

Here's a sample email I sent out this morning. I've yet to hear back but hope springs eternal. As in all such communication, I made sure to include lots of personal details about my history with the influencer, to hopefully tweak Scott's interest:

Scott:

It’s been a couple of years since I had you on the, now defunct, AzureSuccess podcast, but I remember the episode fondly, not in the least because you taught me about African handshakes. I’d apologized at some CSA event for shaking your hand a bit too vigorously and thereby hurting you—you’d had recent shoulder surgery?!?—but it turned out that I had the wrong of it, in that your holding your elbow was actually a traditional sign of respect. I couldn’t help but be flattered, but the key takeaway that day was the realization that I should try to be more-present in my own such interactions; something I’ve strived to do ever since.

I’m writing you today because I’d like to invite you to do another podcast; not mine, but that of my friend John Sherman. John’s For Humanity Podcast focuses on AI existential risk awareness, but from an everyman perspective. I’m quite sure you’d be an ideal guest (in your private capacity!)

I have no idea what your own views on AI risk might be, of course, but given the fact that virtually all of the leaders in the field are signatories of the following statement, including Open AI’s CEO Sam Altman, Microsoft CTO Kevin Scott, Microsoft Chief Scientific Officer Eric Horvitz, Goeffrey Hinton—the so called “Godfather of AI;” the man who literally co-invented invented LLMs, and thousands more, I thought you might want to weigh in.

Mitigating the risk of extinction from AI should be a global priority alongside other societal-scale risks such as pandemics and nuclear war.

It’s fashionable to discount such earnest pronouncements, to cynically suggest that the signatories didn’t really mean it, that AI will automagically turn out be perfectly safe—safe, even though capability spending exceeds safety spending 100 to 1—that our kid’s lives won’t be put at risk, but that’s not my view. I think it’s the time for voices to be heard; yea or nay. Your voice, included…

Louis S. Berman
SquidEyes, LLC; CTO
[email protected] or [phone]
https://www.linkedin.com/in/louis-berman

AI without limits is a ticking time bomb for humanity!
Learn more at https://pauseai.info and https://youtu.be/wFXYQzeftyQto

[twin turret]

**Outreach Tracking Form: Please take 30 seconds to fill this out whenever you reach out to or meet with an influencer, politician, organization, or anyone else!! **https://airtable.com/appWPTGqZmUcs3NWu/pagoraIwzlYm2QwjD/form

Context: we are trying to keep track of all the amazing outreach that's taking place. Having everything submitted in a central database will help enormously with this & help us follow-up more effectively!

[silver prism]

we should add past outreach too, right?

[twin turret]

Yes!! That would be fantastic.

[woven comet]

I am a small (about 100 subs) youtuber and i definitely will work on a video with the resources provided by PauseAI

[rich imp]

Whoah nice! 😁

[twin turret]

Send your YouTube channel if you like! Maybe some of us can subscribe 🙂

[woven comet]

Ooh, it's just a bunch of looser gaming stuff, but I'll make sure to send yall the video before I publish it for review!

[oak hemlock]

I realized today that my wife is a tenured professor (and one-time department chair) at a huge university filled with students I could reach out to. Here's my first attempt:

Dear Professor Payton:

I doubt if you know my wife Set and Costume Designer wife, Marie Anne Chiment, but she suggested I reach out and introduce myself. Marie is a tenured professor in (and former chair of) the School of Theater, Film and Media Arts.

As to myself, I like to refer to myself as a “failed retiree,” in that I left my Chief Technologist role at EPAM (a PA-based 50K+ employee programming consultancy) to have infinitely more fun slinging coding at SquidEyes; a currency-trading startup I co-founded in 2022.

The reason I’m writing you today is because I’m also working to mitigate the risk that superintelligent AI poses to humanity, something I’d like to speak you’re your students about.
I have no idea what your own views on AI risk might be, of course, but given the fact that virtually all of the leaders in the field are signatories of the following statement (including Open AI’s CEO Sam Altman, Microsoft CTO Kevin Scott, “Godfather of AI” Goeffrey Hinton—who literally co-invented invented LLMs, only to ultimately leave Google to be able to talk about AI risk more openly—Bill Gates and thousands more) I can assure you that the risk is not only real, but that it is something our future tech innovators need to become aware of.

Mitigating the risk of extinction from AI should be a global priority alongside other societal-scale risks such as pandemics and nuclear war.

It’s fashionable to discount such earnest pronouncements, to cynically suggest that the signatories didn’t really mean what they said, that AI will automagically turn out be perfectly safe—safe, even though capability spending exceeds safety spending 100 to 1—that our kid’s lives won’t be put at risk, and that our jobs won’t end up on the chopping block, but that’s not my view.
Thanks for your time and attention…

Louis S. Berman
SquidEyes, LLC; CTO
[email protected] or [phone]
https://www.linkedin.com/in/louis-berman

AI without limits is a ticking time bomb for humanity!
Learn more at https://pauseai.info and https://youtu.be/wFXYQzeftyQto

[jolly jay]

"co-invented invented" 👀

LanguageTool is great for errors like that, also available for Microsoft Word afaik since I’ve been seeing you sending docx documents if I remember correctly

[oak hemlock]

Thanks for the heads up

[dreamy flume]

You also said "wife" twice at the beginning.

In the email's final form, the statement on AI risk should be formatted to make it clear that it is a quote.

I noticed that you like the word "automagically". 😄 It's a fun word, but in a serious context I think it sounds dismissive/mocking and could put people off who don't already agree. I think you could get across a similar sentiment with "will just happen to turn out to be perfectly safe" or "will serendipitously turn out to be perfectly safe" or "will turn out to be safe by lucky accident". (Either way, you are currently missing a "to" in "turn out be".)

[silver prism]

since the last people I told you I emailed, I emailed: kurzgesagt, philip defranco, peterson, shapiro, 2 spanish science/ AI channels, captain sparklez, contrapoints, coffeezilla, syrmor

[oak hemlock]

Thanks for the editing help. I'll try to be more careful in the future.

[dreamy flume]

I'm happy to help edit anytime! I'm a better editor than writer and it's something I enjoy doing. I don't mean to come off as critical; keep up the great work!

[twin turret]

That's incredible! Please track here if you haven't already: https://airtable.com/appWPTGqZmUcs3NWu/pagoraIwzlYm2QwjD/form

[twin turret]

Sam is the man. I hope some good comes of this exchange!

When you have a minute, would you mind logging your outreach to Dr. Harris and the other public figures you’ve reached out to, with this link? I know you've done a ton of valuable outreach and it would be helpful to have this all in one place. https://airtable.com/appWPTGqZmUcs3NWu/pagoraIwzlYm2QwjD/form

We’re trying to track all outreach to influencers/officials in order to improve our sense of where PauseAI stands in this vast media landscape. Plus, there will be a leaderboard for those who have done the most outreach 🙂

Same goes for @zealous mulch and @hallow depot - it looks like you've both done a bunch of outreach to influencers/elected officials. Could you take a minute to log your outreach efforts here? https://airtable.com/appWPTGqZmUcs3NWu/pagoraIwzlYm2QwjD/form

(You will see your name in lights on the outreach leaderboard!)

[dreamy flume]

Aiming high! I just reached out to John and Hank Green, and submitted the details of that to the Airtable.
https://docs.google.com/document/d/1SbifPsQGbNJnizRD2Gzx10SX1DvXgz18YGKWJRFjkeY/edit?usp=sharing

If there are any other Nerdfighters here, you should also consider reaching out to Hank or to the pair. They have a lot on their plate, but an x-risk-aware John and Hank Green would be massively OP as voices on this issue!

[silver prism]

I still get surprised that this kind of video from really popular youtubers exists without me knowing at all https://www.youtube.com/watch?v=XlCHEONTqnc 4.4M views, 6.5 years ago. I'll definitely e-mail him, I like to aim high as fuck too lol.

[hallow depot]

https://www.youtube.com/watch?v=20TAkcy3aBY we should contact Jon Stewart/The Daily Show asap

[dreamy flume]

I think this is a huge opportunity: https://discord.com/channels/1100491867675709580/1225072373900906658
(See the linked news thread for a discussion about reaching out to the artists who signed.)

Who here is savvy about how to talk to celebrity musicians, or how best to reach out to them?

[silver prism]

Yeah, I think it make sense to write a message and send most people the same message because the letter seemed pretty friendly to AI and honestly not really promising. I'm not sure how much they would like a pause and how much we can help them with a pause really, but seems worth trying.

fuck i feel like annoyingly pessimistic saying that, should I say stuff like that? The pros I think are: can help us prioritize better + less of a let down if it doesn't succeed?

[dreamy flume]

Yeah, I don't think it's promising enough to warrant the time it would take to craft individual messages for each artist. But I think it is promising enough to warrant the time it would take to send the same message to each artist. Even with a low chance of success, that's a large number of high leverage opportunities, and the fact that they signed an open letter related to wariness around AI gives us just a bit more of a chance, imo.

[twin turret]

"You think we're powerful now, just wait till the AI giants start to stroll around the Earth - and that's coming, you will absolutely see that in your lifetime. You'll probably see that in the next 2 or 3 years, because that technology is moving so quickly that even those who are on top of it can't keep up. So I would say we'd better get our act together, because we could make this world quite the heavenly place, but there's a lot more pathways to hell than there are to heaven."
Jordan Peterson (https://www.youtube.com/watch?v=k0lOL5Q58Ls; 5:42)

Has anyone here reached out to Dr. Peterson? There might be some potential downside - given that he's become a politically controversial "culture war" type of figure - but also some potential upside, given his level of influence and audience size.

I will also note that he’s testified in front of the Canadian parliament before (on an unrelated and admittedly culturally polarizing issue) so he’s no stranger to political advocacy.

That said it would probably be good to proceed with caution, again given the controversies of Dr Peterson. AI becoming associated with “culture war” issues is literally the last thing we need. But I figured I’d throw the idea out there and see if others have suggestions.

[silver prism]

Yeah he has shown to be worried about AIXR

[hazy tide]

Romney seems worried

[noble escarp]

I've recently seen this podcast channel because they focused on AI for some episodes. The host is based in California, and since Liron is also, I thought it could be interesting to try something.

Cons

• Thumbnails suggest it's for a catastrophist audience: "The world is ending in 10 minutes" kind of vibes.

• Tom seems to be a techno-optimist with an accelerationist mentality, and he has only had techno-optimists (mostly entrepreneurs) discussing AI on the podcast so far. So the line of questioning may be hostile.

• It's an in-person podcast, so someone has to be physically present in the studio. This is more difficult to plan than a Zoom call.

Pros

• Big podcast with big outreach; other episodes on AI are averaging 200K views.

• Tom seems interested and open to talking about AI.

• Reaching out to techno-optimists with an accelerationist mentality and finding common ground by discussing risks and possible regulations would be great.

• Geographic proximity to Liron, who has already appeared on multiple podcasts, is very well-spoken, and is a prominent figure of PauseAI.

Is there somebody I should contact at PauseAI to see if it's possible, or should I try to reach out to Liron and Tom myself?

https://www.youtube.com/@TomBilyeu/videos

[hazy tide]

@reef warren is here if you want to talk to him

[noble escarp]

Oh great, thanks, I didn't notice!

[reef warren]

Howdy, i haven’t seen this podcast but always happy to do that kind of stuff if I get an invite 🙂

[noble escarp]

Great! I'll reach out to them and come back to you with their answer, thanks!

[tame roost]

Hey everyone, I've posted over in the #protesting-📢 channel and was pointed this way. After reading through the history here, it seems that influencer contacts have been fairly unsuccessful so far. I've been thinking of a broad framework that we could work within that might be more effective than individual contacts, I'm going to just crosspost the meat of it here

I've been thinking about how effective we could be if we were to attempt to leverage AI-safety-aware content creators across communities on YouTube and other platforms to push a single massive protest in DC or something of the sort. I'm thinking such a drive could be structured like this:

1. Get in contact AI-safety-aware creators like Kyle Hill who have previously acknowledged the monumental danger we are facing and ask to do a collaboration with them. We should approach with a fairly concise and understandable message, gather the best examples of the crazy Ai tech that exists now, and generally make this as easy as possible for the creators to join in.
2. Get everyone who signs on to put together a video and set a single publishing date.
3. These videos ideally cover the main concepts surrounding AI safety like the alignment problem, the arms race dynamics, conversations about incentive structures, concerns around art and job loss, and the destabilization that even current-level systems are going to cause from misinformation, etc... These should be tailored for the creator's community and should try to be as accessible as possible.
4. The videos should also include a direct call to action announcing a protest date and location alongside collaboration with PauseAI (and potentially reference our earlier protests).
5. We need to be ready as a community to address disagreements that people have in the comment sections on these videos when they launch. All too often on AI safety videos I just see a wall of e/acc BS with absolutely no pushback.
6. Get media attention using our contacts that have been established.
7. Encourage other creators who may not have been involved in the initial push to join the call (get a 30-60 day window between intial launch and protest).
8. Find ways to make this event an appealing thing to attend. See if we can get famous people in the AI space to come and talk.

Ideally, having a set time and well-put-together website or frontpage for the planned protest would give some more legitimacy to it. I also imagine that if we were able to send a list of influencers who had signed on, that would help convince others.

This way, rather than just approaching them about making a specific video, we're more approaching them about promoting an actual thing that people can do. I'm not deep enough in the creator space to have a good feel for whether it would make more sense to try to get independent videos for this or whether it might make sense to purchase sponsorships

[hazy tide]

Can you help lead this, then? We will be happy to follow.

[tame roost]

I would be happy to help lead the influencer outreach for this, obviously though we'll need to coordinate with the leadership for organizing the actual event and dealing with paying for potential sponsorships

I'm still fairly new to PauseAI so I'm not very familiar with the organizational structure and way of doing things

[silver prism]

check #1239733713554898974 for a potential useful tool

I think this idea seems pretty good but it's something that could take a long time, there's a lot of unknowns.

And I think that sponsored / not sponsored will be the most important one.

[tame roost]

Did we ever figure out the costs of this? I think it might be important to start by targeting influencers who have already posted something about AI. We might be able to just avoid the cost by taking submissions from the community here as a whole and compiling a large list

[tame roost]

It seems optimal if we can do it without sponsorships, but this is also fairly outside the realm of normal for most YouTube channels. Sponsorships would definitely be the surefire way to collaborate with the channels, but would need significant capital to get off the ground on many channels simultaneously. Additionally, I think our message would be better spread if it comes from the creators themselves and not through a sponsor slot that most people will skip

[silver prism]

it says Tuesday so nope

[rich imp]

Hi Max, amazing! I can help here, also with funds. Love the energy!

[nova nest]

The Influencer Intelligence database is £3000+VAT for a one-year licence. This is a reduced offer because of being a non-profit if we accept the offer soon. I just had a demonstration, but they could do more demonstrations if others want to see.

[tame roost]

I'm at work so can't talk a lot, but how was the demonstration? Is it essentially just a giant database of influencers that can be filtered? Do you think we could accomplish approximately the same list for our specific case just by crowdsourcing influencers via our community?

[nova nest]

Yes, it's a giant database of influencers. We would have to play with it to find our target markets, but possible searches include "influencers who are parents with an interest in technology" or "influencers with an interest in climate change and technology". One can search influencers for a search term like "AI" and then be referred to all their past messages on AI & analysis of those Instagram posts (or Twitter or YouTube etc).

I attach a PDF and summary of it. I don't know if it's the best use of £3000+VAT, but do think it would expedite our influencer efforts. To answer your question, I don't think we could crowdsource exactly equivalent value.

Another consideration is that this is the cost for one or two licenses. I figured we could share passwords rather than pay a higher subscription fee for more licenses.

[mint shuttle]

I like your idea @tame roost of content coming organically rather than creators being paid to do like a formal sponsorship where they have to mention it a certain way -- it might seem less believable and it's pretty rare for causes to advertise this way (it's mostly brands that have a financial stake that do this).

outside of paying creators, we are banking on people finding something else in it for them -- like helping them to get more followers if they are a small creator and featuring them on our social media helps them to gain viewers. or else just a personal feeling about the issue, maybe they are already concerned about AI, or they feel it's a good topic their followers would be interested in.

there are so many creators we can find organically -- just go on a platform of choice and write to those you follow already, or do searches (like for artists, computer developers, etc.) and contact anyone who shows up in the search.

this issue came up in last night's welcome meeting (didn't realize it was already a project) and @stable carbon and I were talking about it. I made this first draft of a potential way to reach out to influencers organically that any member could send to influencers they follow or find, based on templates I saw for influencer reach-outs.

https://docs.google.com/document/d/1WV9Gyf1Qu3TW95aDLlpOo0Ldwmj2gHA3CeO1eVtR0AM/edit?usp=sharing

[mint shuttle]

I think the paid influencer database license might be more for commercial brands who
-need influencers of a certain size and reach to make a paid partnership worth the money (it's advertising spend)
-might not be people who follow that kind of influencer themselves and need fast intel on it for selling a product in a particular niche, such as young parents (our market for the cause is pretty much anyone)
-are able to offer a paid partnership to make it worth it for larger influencers to spend their time on something and compete with other brand partnerships they are being offered (I don't know if we have budget for this or if it would be even be a good look to do paid advertising for a cause)
also it is unlikely the license filters for influencers' opinions on AI, something we could find through searches

If we are instead just asking people to cover the topic out of caring about the issue and/or wanting to get more followers we can probably find influencers ourselves through keyword searches? maybe it would be good in that case to have a database of who has been contacted so people don't get spammed if they made one critical video about AI. but could also branch out to adjacent areas like artists, anyone whose job might be threatened, etc.

[mint shuttle]

One of the influencer marketing campaigns I really liked was for Dossier fragrances. They contacted tons and tons of nano (very small) lifestyle and beauty influencers and offered them two free fragrances (value <$30) for making a 5+ min video about their thoughts on the products. (Dossier the company was probably approving the videos and giving them pointers of what to cover, but it came across as really genuine because the influencers got to say whatever their thoughts were.) I ended up seeing tons of these videos from different people's perspectives, and it got me to try the product.

"Dossier knew from the beginning that influencer marketing was the right marketing strategy for them, “we’re strong believers in influencer marketing. We’ve been utilizing influencer marketing long before it was a fancy term. Last year, there were about 15,000 pieces of content that talked about Dossier in some form or fashion. So when I say we’re believers in prospecting, we really are.” https://gomalomo.com/blog/selling-and-marketing-perfume-online-dossiers-journey-to-success

If we have money to spend on influencer marketing, maybe we could offer influencers (not those who we think would partner with us either way, but maybe those with a slight overlap to our cause but followers who might not know about this issue yet, and who have at least a certain # of followers) a free PauseAI t-shirt or cap (that they could wear while making the content, and afterward) if they make a video talking about AI dangers and telling their followers to check out our nonprofit and cause?

In contrast to traditional media journalism, where it would be mostly tech journalists covering this, influencer marketing where we provide swag could bring the cause to new markets of followers. Like general activism, artists, musicians, writers, job search, environment, animal rights. would need to get approval to make sure the influencers had the right # of followers and right brand to be associated with us.

[rich imp]

We have funds available for things like this! I think we'll need an influencer team. Maybe @mint shuttle would be interested in leading it?

[mint shuttle]

I would be interested.

[noble escarp]

Reached out to the Jordan Harbinger show, he's not interested in having a guest from PauseAI on the podcast for now, but he offered a possibility for a paid media slot. This slot can be specifically during a podcast episode about AI. Would that be okay? If it is, what's the acceptable cap on pricing? https://www.youtube.com/@JordanHarbingerShow

[rich imp]

IMO we should keep the option open for paid ads, but I'm pretty sure most podcasts have this possibility. It seems likely that there are some free options where we'll get far more screen time.

On tiktok I can get 80k impressions (couple of seconds of attention) for about 50 euros. A slot in a podcast should probably be more valuable per view.

[silver prism]

I left some comments in the Michelle documents, I like it

About the influencer intelligence... I'm not really sure, I think there could be small creators that could be interested. Because lots of big creators who has shown interested on AI has been contacted and there has been no response from them

Paid ads I think should be our last resort? I don't know if it's weird to ask for a free collab and some months later a paid collab, but it shouldn't. I think that we should be able to find small influencers interested in doing it for free

[silver prism]

also it is unlikely the license filters for influencers' opinions on AI, something we could find through searches
but rochelle said:
One can search influencers for a search term like "AI" and then be referred to all their past messages on AI & analysis of those Instagram posts (or Twitter or YouTube etc).

[mint shuttle]

my mistake! Well, maybe try just finding influencers ourselves to start (we can use searches and look at people's follower numbers, their styles) and consider the database later if needed? we could probably get team members who are familiar with the various social media channels to help with finding influencers.

if we have a limited budget on this, we might be able to get around 100 influencer collabs (if we can get people to agree to them) for the same cost as one year of the license. try it for free first and see if we need the paid license later?

@silver prism thank you for your comments on the doc. Good points.

1. As @silver prism said, we really should get our social media looking good prior to reaching out. Copy at least the really good videos across all our platforms, if not all of them. I can work on a list of which videos on each platform still need copying to other platforms.

2. Fourthwall gave me these instructions for offering a one-time promo code for free merch up to a certain limit as well as free shipping. Are we able to offer this on the PauseAI store so that influencers could order themselves a free swag item without having to give us their home address? https://help.fourthwall.com/hc/en-us/articles/13331335603227-How-do-I-create-a-promo-code
   (I don't see anywhere on our merch store for a promo code but Fourthwall seems to offer the capability for it)

3. Maybe we also try for fully free collabs as well (like approach people who already seem sympathetic to our cause carefully, suggesting they interview a member) but at least while we are figuring things out, a small swag seems like a small cost to pay for what might end up being really high-quality advertising. giving people anything at all can make them more open and collaborative and make us not sound like jerks for asking them to do something without offering them anything in return.

[rich imp]

https://twitter.com/skdh/status/1793330652059766893?t=CKS52hRBc1DgGSeE96H_Dw&s=19

Sabine hosaenfelder is slowly but surely starting to get it!

[hazy tide]

I want to contact Sabine Hossenfelder - any thoughts before I do it?

[hazy tide]

You'll be surprised how effective emails are

[mint shuttle]

Oh cool, here is another influencer database (Modash) but it costs a somewhat less, and it has a free 2-week trial available.

The person reviewing it says it helps you to find the influencers' email addresses (which can be hard to find otherwise), which helps to stand out and not just be sending a DM.

https://www.modash.io/

https://www.modash.io/pricing

[rich imp]

Carefully read her tweets, check out her videos

show her that you watch the channel and you know her takes and how they've evolved!

Spend at least 30 minutes on research, probably more

[silver prism]

#media-📺 message plausible sympathetic channel

[rich imp]

New video from Sabine: https://youtu.be/pZVngnr8KN4

She's genuinely worried about AI. Not about alignment yet, though

[hazy tide]

https://sabinehossenfelder.com/contact/

[dreamy flume]

I think we can confidently say that the middle-of-the-road position for the scientific community on this subject is "Hm, yeah, there's some chance this could all go horribly wrong. It probably won't, at least not that way, but yeah, that could happen."

[woven yarrow]

In my experience, usually followed by "this is the specific adjacency that I'm distracted by."

These can be useful for conversation, but they do tend to deflect thinking about "how should we and I react, given there's some chance this could all go horribly wrong but the detail seems unclear."

[rich imp]

Sabine was slowly going towards Yann's side (as she dislikes elon and they both had a fitty on X), which now even resulted in her talking nice about Yann's recent PauseAI ridiculization: https://x.com/PauseAI/status/1797587541954117719

PauseAI ⏸ (@PauseAI) on X

@skdh @the_yanco @elonmusk @ylecun Yann is strategically using language to dismiss concerns.

"Apocalyptic cults" is a frame he uses to dismiss worries surrounding x-risk. He's saying we're a cult, which is weird accusation for a decentralized activist movement.

He also uses the term "Fringe doomers" a lot, which

Twitter · Jun 3, 2024, 11:14 AM

[hazy tide]

Still seems to be worth trying to talk to her in good faith

[hazy tide]

I haven't written to her yet. I'll try to get around to it tonight.

[tame roost]

I just sent Sabine an email cautioning against Yann and offering to open a dialogue either personally or with PauseAI as an organization

[hazy tide]

Thank you, I'll hold off then. She scares me for some reason.

[tame roost]

I like her particular brand of sarcasm and no-nonsense approach to science communication, especially coming from a physics background myself

I think she will likely become a very important voice in the AI safety movement given time and proper communication

Her content on climate change being worse than the IPCC talks about is one key indicator for me

[hazy tide]

I keep wondering why truly intelligent people would so casually dismiss risks, since you don't even need anything special to get Christano's "Failure feels like this."

[tame roost]

I've done a lot of outreach in the physics department at my university including an undergraduate research seminar titled "Why Everyone Should Care About AI Safety: A Physics Perspective," so I've had a decent amount of experience with talking to "truly intelligent" people who initially do not buy the risks.

I think the biggest reason is that AI is still a fairly niche field in terms of actually understanding the fundamental issues with our current approaches such as the true extent that we do not understand these systems, understanding how they primarily just are scaling with more compute, etc.

Usually if I sit down and talk with someone for 40ish minutes, that's enough to get them on board. It just takes someone actually walking them through the steps of understanding the alignment problem, and many smart people are so specialized into their own unique fields that they haven't spent the time to do that on their own

The corrigibility of people when I actually sit down and talk with them is one of the reasons I do have some level of hope that we can actually win the public consciousness. We need to focus on spreading the counter-arguments to all the stupid shit Yann says, and explaining the root issues for people.

I have found that "normal" people really do understand these issues when you put them in terms of creating new competitors for humans in darwinian terms beyond anything we've ever had to compete with before

[tame roost]

https://youtu.be/QO5plxqu_Yw?si=WMZmqXz3_dmC7bWU

Yesss Sabine!

I take full credit for this lmao (she has not responded to me)

[glad tinsel]

Maybe try got get influencers to pick up on NVIDIA saying they'll make it easy to replace humans in all customer service jobs: #1247528017220079757 message

[glad tinsel]

High profile influencers contact messages deserve more attention, because first impressions matter in their case. Otherwise, well written templates would make it so much easier to get started that it should be worth the extra effort. More on message templates

[rich imp]

The Hossmeister has awakened! https://x.com/skdh/status/1798322646540529957?t=jeIjTNy0ETxBdCOwxwfZzA&s=19

Sabine Hossenfelder (@skdh) on X

AI becoming intelligent and AI developing its own values and being able to act on them are two different problems. The former -- intelligence -- is likely to precede the latter. I don't think it will be long and I think not enough attention is being paid to the most obvious

Twitter · Jun 5, 2024, 11:55 AM

[woven yarrow]

"If you don't understand why superior intelligence is scary, you've read too many touchy sci-fi stories in which human irrationality saves the day. But humans didn't come to dominate this planet because they're sometimes irrational, they came to dominate despite of it."

Nice phrasing.

[hazy tide]

Now if she will go through the rest of the discovery cycle

[rich imp]

right on time, because the day before sabine was simping a bit for yann (because it was musk vs lecun day and sabine doesnt like musk)

[hazy tide]

We need a list of influencers and contact team.

[rich imp]

good job for the new #outreach-team

[dreamy flume]

I sent Sabine a quick thank-you email for discussing AI, including its risks, and offered to provide additional academic resources in case she happened to be interested.

[glad tinsel]

Here is NVIDIA's own wideo about CEO keynote at Computex 2024, link goes to timestamp 48:57, where he really goes into the "We will make it easier than ever to replace any customer service worke" spiel: https://www.youtube.com/live/pKXDVsWZmUU?t=48m57s

[tame roost]

I like how he says "just tens and tens of millions of customer service (jobs) can now be augmented with AI"

He leaves out the word jobs and then says augmented as if he isn't presenting a model designed to completely replace them

[hazy tide]

I think that if a short can be made of it, we can publicize it

[tame roost]

I think youtube has a built in clipping feature, let me see what I can do

[tame roost]

Should I cut it off after the quote I mentioned or let him go on to explain the NIMS?

[hazy tide]

How long would it be if not cut off?

Keep at 60 seconds or less

[tame roost]

There's no nice ending less than 60s

I'll cut it after the quote

Here's the clip, I don't think this is exactly in the form of a short. I need to figure out editing software at some point here to properly make videos

[hazy tide]

@opaque wing perhaps you can chime in a second on this?

[tame roost]

I'm adjusting the bounds on the clip a tad

https://youtube.com/clip/Ugkx04ik3aQg1k4gpkuiiiIMnb519fWZLjYG?si=meHVSCWsu0v1GMb4

Here's a slightly better cut (just cleaned up the bounds)

[jolly jay]

Are clips recommended to people or are they just meant for sharing?

[opaque wing]

Oh yeah, we could turn this into a cool clip!

I'm so incredibly busy today, but I can try to get this edited. If you guys can send a bunch of screenshots of workers being replaced by AI, I can add them to the video! That would help me a lot

Eg:

If anyone wants to make a google drive folder and add the screenshots, I can edit the video.

[hazy tide]

Awesome, that sounds like something Max or das can quickly do in terms of drive folder?

[tame roost]

I'll start a folder and add what I can find to it. I'm going to be driving back to my parent's today for a dentist appointment tomorrow morning so that's a 5hr drive lol

[tame roost]

How would you like me to note videos?
I can screenshot like attached, just thumbnails, links, etc...

https://drive.google.com/drive/folders/1gEqeUOGB9g_qUxOM_jUQKTnRwyG2RxpM?usp=drive_link

^Drive folder link

[tame roost]

https://docs.google.com/document/d/1f_OrL_TvI1hKQpD24c3FHUEKJRXwX5HhUTN_tSTq348/edit

Here's a compilation of what I could find between videos, articles, and reddit posts

[opaque wing]

Ideally just screenshots of news articles - gives it more credibility

[tame roost]

Ok, well I have a good few articles linked and screenshotted in the doc listed above

There are also videos and some reddit/twitter posts/comments

[glad tinsel]

Clips are only meant to be shared, apparently, so they are very different from shorts.

Ideally, someone should

1. make a longer video commenting on that plan, so that this part of the NVIDIA video, perhaps with more comments as a text overlay, could be included as a part of that longer commentary video (thus hopefully avoiding copyright takedown(although Nvidia might just cheer free publicity. They probably hope for it... ))
2. Then create a short from that commentary video.

[dreamy flume]

Can you move this folder into the Projects folder of the PauseAI Public Drive?
https://drive.google.com/drive/folders/1Ga3aqkO-uXoDNIlhbZCgzt1QA9fMtnhQ?usp=drive_link

We could also create a Job Loss Compilation project thread for this effort.

[glad tinsel]

Related to refuting individuals:

While I am using Matthew Berman as an example, this applies to many more influencers.

While I do think that articles / posts / videos about "why smart and mostly correct person x is wrong about risk y" might make a lot of sense in some cases, such as if convincing them seems totally hopeless, in some cases it would be better to try to engage them in some sort of dialogue, and making an extraordinary effort at finding best ways to convince them.

Matthew has almost 300 k subscribers on YouTube and makes a lot of content about AI, most of which is positive, and a lot of it is tutorial type stuff or other hyping. But he does acknowledge many risks, expresses worries about them and criticizes bad behavior by AI companies. He also apparently is kind of open to having his mind changed, but so far is not impressed (top rated comment on LessWrong post about new MIRI communication strategy.)

While he says that he has engaged MIRI / MIRI aligned people a lot over the years, maybe all it would take to convince him that one of his key assumptions is wrong would be to present the right metaphor for illustrating why some analogy is not suited to reasoning about AI, to make it "click" for him, or something like that.

I might be reading too much into that individual comment to gauge what are his key questionable assumptions / reasoning steps, but, it seems some of them are:

• training on human data would make them in some ways like humans
• feedback mechanisms would instill human values in them
• maybe skeptical about AI having values not reflected in outputs
• maybe thinks AI would not have "meaningless' values
• says his P(doom) is " more like < 20 %", does not think that warrants a halt.

[glad tinsel]

Related: Project to compile data and guides on what what works best against various failure modes in reasoning about AI.

Also relevant: If you think "preventing possible AI doom requires giving up too many certain / probable good things x, y and z", that creates a really strong emotional barrier / disincentive to accepting even "AI doom is a serious risk" let alone "risk of AI doom requires drastic action". Because of that, it might be a good idea to preemptively present the idea: "we think there are many ways in which we could prevent AI doom without giving up good things x, y and z, such as..." and even better: "and they could also help with these other risks too, also without giving up all good things"
Project: Responses to opposition of specific mitigation measures. For example, in Matthew Berman's case, ideas on how to control dangerous development without killing open source.

[opaque wing]

Quick update that I'm working on this. Super busy week as I said, but I'll get this done.

[hazy tide]

Let me know if I can help.

[mint shuttle]

great video! and that's a great article he's reading. and he is right about Luddites having a point.

You are right that Vaush is really controversial, but cool to see him going so deep into this!

With regard to influencers,

1. there's official PauseAI influencer outreach for partnerships or influencing people where we have to think about our image
2. and then there's also just the possibility of writing to influencers you follow as an individual, not bringing up PauseAI, just sending them an article or something in case they take an interest. Could be a lot of streamers who might want to talk about it, it's certainly an interesting topic. I am starting a team for doing this kind of thing (Recruiter team). It should be up and running soon if you are interested and have anyone you like to watch.

[hazy tide]

https://x.com/skdh/status/1800113740047945887

Sabine Hossenfelder (@skdh) on X

wondering how many academics realize that scientific publishers have almost certainly sold off copyrights on their papers to any AI company willing to pay money and that the authors of those papers will see nil of it

Twitter · Jun 10, 2024, 10:32 AM

[rich imp]

I've reached out to Digital Engine (10M+ views on youtube)

Aligned channel!

[hazy tide]

Alas her latest video is just "no way AI can remake the world economy so quickly"

https://youtu.be/xm1B3Y3ypoE?si=3846Q6S1LKHc2O7x

But only if you have written for SB-1047 🙂

[rich imp]

Lots of bad takes here. Sad to see it.

[woven yarrow]

Yeah, a bit disappointed.

Engagement with the actual arguments seemed lacking, relative to theories re prediction veracity.

[rich imp]

I suppose this is what happens when people who are specialized in a field (e.g. theoretical physics) start to feel cocky after getting a bit too popular.

[woven yarrow]

We're all human. Sabine is smart. Guess: respectfully earning trust seems like the move here.

[rich imp]

agreed

[nova nest]

Actor Joseph Gordon-Levitt is on board 😍😁https://x.com/EncodeJustice/status/1801730211446407202?t=Jfh7frCDGfpRD6Fy5QogJw&s=19

Encode Justice (@EncodeJustice) on X

So... we just got to do something pretty cool 😁

We sat down with our childhood icon Joseph Gordon-Levitt (@hitRECordJoe), actor and AI 2030 signatory. We covered his interest in AI risk, the AI ethics/AI safety divide, Biden and Trump, and more.

Twitter · Jun 14, 2024, 9:35 PM

[rich imp]

Nice, I met sneha and sunny today BTW!

[rich imp]

Both Channing Tatum and Scarlett Johansson made remarks about AI:

Scarlett spoke out on need for legislation. “We're all waiting and supporting the passing of legislation to protect everybody's individual rights.” "This highlights how vulnerable everybody is to [AI] and how little protection people do have - if any - of their work in their likeness,"

Channing Tatum: "So many people are going to lose their jobs just to an AI that can do so much," he told reporters. "I just can't believe that's not even a bigger story than it is right now."

Thanks to @regal sky

[rich imp]

https://youtu.be/UcEyfQ1I8jg?si=pw-bILAEfqOI5gUl

Sabine Hossenfelder shares her takes. Disagrees with both Yann and Geoff. Thinks it's not about control but resource competition. IMO she seems to get it! Or she's close!

[rich imp]

Reached out to her again

[regal sky]

Here is video of remarks made by Scarlett and Channing. https://www.youtube.com/watch?v=F1OsI7X4m14

[humble breach]

Might be worth tagging people using the social media accounts to get their attention instead of using private methods?

[rich imp]

Have done that too!

[rich imp]

@thorny sun wants to reach out to some influencers, get them to support PauseAI!

[regal sky]

Myself, Diego, and a couple others had call on this very subject last night.

[dreamy flume]

Has anyone reached out to Nathaniel Beardsley of the YouTube channel "Generic Entertainment"? (He made this video last year. Very entertaining and Pause-pilled.)
His Twitter account: https://twitter.com/nathbeardsley
(And he has a small Discord server, but I didn't want to join it just to spam.)

[rich imp]

I'll send him a DM!

[rich imp]

And he replied! My message "made his day" and he's very supportive. a real G!

[silver prism]

I'm compiling different kind of contacts for preparation for warning shots (although it is useful for other cases), and in between them there are the influencer emails that we reached out/ could reach out to https://docs.google.com/spreadsheets/d/1WIAUzxS23nkSJWe3jgLZFYYM7zNAjmM-s-eD4dOn31g/edit?gid=0#gid=0
If you don't have access to this doc and you want it ask me. I don't think it would be good to have the doc super public.

I don't think I put everyone there and the hardest job is to actually look up their emails again lol, so I would appreciate a hand

[regal sky]

Myself, Felix, SP, and Holly get together virtually on Tuesday nights (tonight) at 5 Pacific. Please reach out to SP or Felix to be added to our group.