I strongly support Holly and Pause AI US!

@ampdot True. IMO there's not another consistently "high P(doom) AI x-risk" outlet communicating in a mainstream way right now.

(I guess the closest thing right now is either explainer channels like Rob Miles, or podcasts that discuss AI issues where some of them are high-P(doom) guests.)

PauseAI member here, amazing organization and the Discord community is great - https://pauseai.info/join

@michaeltrazzi You're right, I've been doing a lot of Liron Reacts lately for a few reasons:

1. It's relatively easy for me to crank these out: The structure is predetermined, I just go through an existing piece of content

2. I can have more prominent people featured in my content and dive into their positions in detail, without having to convince them to debate me, good for attracting new viewers and YouTube keyword SEO

3. It often lets me give higher-quality arguments and explanations than in a debate, because I don't have to argue in realtime (though I can mitigate this issue for debates by doing a post-debate recap episode, which I did after debating Robin Hanson)

4. I'm getting frequent listener feedback like "Yesss I was hoping you'd review this", because many people who follow the AI media space are in a similar position to me, feeling frustrated that the doom issue gets ignored / misrepresented / badly argued against on popular podcasts multiple times per week, and they end up hate-watching these other episodes, but now they can love-watch Doom Debates instead (or in addition to), to follow the conversation from a "taking doom seriously" viewpoint

Because of these factors, in the short term, I'll probably do something like 2 reaction episodes or other solo explainer episodes per week, plus 1 debate episode every 2 weeks. (FYI I'm planning for a debate episode next week around SB 1047 where it's me moderating two other people debating pro vs. anti.)

As we progress from the humble origin of being a niche community of passionate doomers looking out at the larger discourse, to becoming a forum that increasingly-prominent people come to participate in, the best way to create value is to raise society's ability to have high-quality debates about AGI, similar to that one AGI MunkDebate last year, though hopefully even better quality than that. That's the endgame I'm targeting.

Along the way, it's always a continuum where the more high-quality guests we can attract who are willing to debate their position, the more we can ditch reaction episodes, because we can go straight to the source. (Case in point, my David Shapiro Part II reaction episode was originally supposed to be a debate with him until he changed his mind.)

• the production quality is quite good on your side, though say when you're reacting to david shapiro he his only inside a circle (i guess for copyright issues?) which I think could slightly be improved

I think that particular issue is just a weird isolated case where David has a weird-looking video that I clipped unaltered, but thx for the feedback and in general I will keep improving the production quality :)

I guess last conflicted thoughts I'd have are on the theory of change. If say more david shapiro folks watch your reacts / debates, or more generally people do understand the basic arguments for why "doom" is likely, what do you expect them to do as a result?

I expect listeners who are interested in the content of folks like David Shapiro to hear both of our perspectives (David's and mine), and hopefully update toward thinking that the "taking AI doom seriously" perspective makes more sense.

I don't consider it my goal to change the mind of the people I'm debating against, because I don't have high hopes that such a goal is achievable in more than a small fraction of cases. Nevertheless, I expect to act very similarly to someone who seeks to change the mind of those people (e.g. use Double Crux as my debate technique rather than rhetoric), because part of what I aim to do is be a social institution for high-quality debate, which starts with modeling to viewers what high-quality debate looks like, and continues into being a Munkdebates-like forum that referees public debates.

My understanding is that your audience would most likely not be that technical, so the goal is potentially to have people more concerned about AI risk to say shift the public opinion on AI legislation, or do call to actions for the PauseAI movement? If so would be good to have data on that (eg a form on how much their perspectives have shifted or how many PauseAI discord signups from your channel)

That's right, it's an important success metric that this has an effect of increasing support for things like AI safety legislation and grassroots PauseAI protests. I will make sure to run surveys on that, measure referrals to PauseAI discord signups, etc.

Thanks again. I hope you and others keep the questions & comments coming.

@michaeltrazzi Thank you for the thoughtful feedback and questions! You clearly have a lot of relevant context as a fellow podcaster, and you're helping me make this project page more informative.

- framing it as a "doom debate" or a confrontaional "doomers vs non doomers" which could potentially cause harm (making the movement be seen more like an advesarial apocalyptic cult rather than a technical field)

Fair point about the name and framing. Another valid criticism of the name is that "doom" is perceived as an unserious outgroup-word for serious policy discussions. My thinking is that until I get to the next order-of-magnitude growth target, my main focus is inspiring passion for the project in myself and my early audience. I find the current name and framing authentic and catchy, and I feel like a lot of the audience is vibing with my overall approach, so I plan to reconsider how to take the next step of the journey at the next order of magnitude.

For instance, let's say we hit the milestone of 50k+ listens per episode, and I'm now optimizing the show toward having regular high-profile guests. In that case, it may be time to rebrand to something more serious-sounding like "AI Safety Roundtable", while bringing along the existing audience that came from earlier iterations.

- one of his debate with Guillaume Verdon (around his startup) might have been counterproductive (he repeated the same question about "inputs / outputs", which I think was a fair question, but this didn't signal to the audience that he was engaging with the actual plan from Guillaume, and was criticized on twitter for that)

Fair enough, I consider that my worst debate and I think you'll find that every episode of Doom Debates is better.

How many listens are you getting in the "views and listens"? Is it 1k+ views + listens in average? What is, say, the average watchtime per episode? What is the average view duration (in %)?

YouTube is the only platform where I'm really doing numbers, reporting over 1k views per recent episode. Substack reports about 60 podcast downloads per episode, and Twitter reports thousands of "views" on my tweets where I post clips and advertise subscribing to the show, neither of which amounts to much, yet 30% of poll respondents on my YouTube channel said they also listen in podcast form. So I figure I can add at least 10% to the YouTube view count to get "views and listens".

YouTube analytics:
25,600 total views since Jun 19 launch. Average watchtime per episode is 13:31, about 15% of an episode length.
5,800 total hours watched.
In last 28 days, there have been 8 episodes averaging 1.4k views per episode, and 70 likes and 83 comments per episode.

An important positive indicator for me, besides the relatively high and growing view counts for a 2-month-old channel, is the passion and enthusiasm of the audience engagement (i.e. likes and comments). This is something I feel qualitatively in the genuine interest and compliments of the commenters, but also can be quantified: Dwarkesh has 218,000 YouTube subscribers compared to my current 640, and his videos have 10-100x more views than mine, yet for all but his top 20% of his videos, his comment counts are actually in the same range as mine (~70 per video).

What are these "obvious" low-hanging fruit growth tactics?

The most obvious way to add a multiplier on the organic growth rate that comes from producing episodes is to post each episode's content and assorted clips across YouTube, TikTok, X, LinkedIn, Instagram, Facebook, and various forums and Subreddits where it's relevant.

Plus, the topic of AI risk is growing increasingly mainstream. So I bet the potential audience who wants to consume my content about AI x-risk really is in the millions.

How did you end up with "in the millions" here?

Rob Miles is a similar kind of mainstream communicator in the same space as me, and has 250k views per episode. AI news is the hottest topic in tech news, and the majority of random respondents in various surveys say they're worried about various serious threats from AI. I figure millions of people are interested in tech news, and millions are interested in learning about whether we're all about to die.

The viewers-per-episode metric will soon hit tens of thousands

What data supports that? What do you mean by soon?

I'm aiming to hit 10,000 subscribers by end of this year. It'll require one or two extra-popular episodes. I see many opportunities to create such hit episodes, most straightforwardly by creating thorough takedowns of popular names in the AI non-doomer space, such as Marc Andreessen.

Every time I make an episode debating or reacting to content from someone who has their own related audience, it helps get more subscribers, and sometimes disproportionately so (e.g. David Shapiro has 200x more subscribers who have recently been coming over to my channel since we talk about similar issues and now I'm reacting directly to his content).

I expect Doom Debates growth on YouTube will be faster than most channels because the issues I talk about are unusually engaging and polarizing, and many of my subscribers are aligned with the show's mission, and are motivated to share and help it grow. For the next 10-100x growth, I'd guess it plays out as a compounding exponential growth / snowball effect where the algorithm keeps promoting my videos more, rather than hitting a plateau.

The most likely failure mode is if my content doesn't resonate with a large audience, making it hard to get beyond 25k+ views and listens per episode, such that I give up on the 250k+ views per episode which I currently think is a realistic long-term goal.

When will you decide whether you're in that failure mode or not? Why did you decide on 250k+ views episode as a long-term goal? What makes you think it is a "realistic long-term goal", and what do you mean by "long-term" here?

If it's stuck in the 25k range for 6 months despite creating various types of content that I think should be relevant to a large audience, I'll be pretty surprised. I would check if the average level of engagement and passion in the comments is similar to what it it is now. If not, figure out why not and make sure I'm still communicating with an appealing mainstream format. If yes, try other growth channels since it could be specific to one platform's algorithm.

If the viewership goes 12 months without a 50% yr/yr increase prior to the 250k subscriber mark, then the project will not have gone as hoped & planned. I may work on it less since the ROI is lower than expected, or I may keep chugging along and be open to serving a smaller niche, especially if I'm creating episodes that serve as a high-quality reference for various topics.

Thanks again for your thoughtful questions and overall encouragement!

This is amazing. A proven team, doing really important and cost-efficient work, showing tons of results.