First things first, I’m pro experiments so would be down to experiment with stuff in this area.
Beyond that, seems to depend on a couple of things:
The details of what inviting viewers to LW would look like.
What the LessWrong team thinks is the best use of our time.
The Details Matter
LessWrong currently has about 2,000 logged in users per day. And to 20-100 new users each day (comparing the wide range including peaks recently). If the numbers of viewers wouldn’t change that much, perhaps +10%, it wouldn’t be a big deal. On the other hand, if Rational Animations was wildly successful and driving several hundred people to create LW accounts a day (with flow through to posts/comments), that’d be a big deal. That would require work to handle.
Another big question is the level of sophistication of users. If you’re sending people to engage with more advanced stuff vs more intro content, that’s a big difference in how I expect them to relate to the site.
Should LessWrong be providing intro material and answers?
Even before you posted this, I’d been thinking about this question. I think it terms of knowledge and skill, we’re well positioned to provide a web resource of intro Alignment stuff. It’s not quite the core thing of “create a thriving intellectual community that can make progress on the important hard problems”, but might be a worthwhile enough opportunity that we should put time into it. I’d already been planning to do some work on our wiki system with an eye towards having more intro AI content for a broader audience.
So maybe. Maybe we should leaning into this as an opportunity even though it’ll take work of both not letting it affect the site in bad ways (moderation, etc). and also possibly preparing better material for a broader audience.
Moderation Costs
To provide some insight: on the margin, more new users means more work for us. We process all first time posters/commenters manually, so there’s a linear factor there, and of new users, some require follow-up and then moderation action. So currently, there’s human cost in adding more people.
It might be possible to build tech to reduce the marginal cost of new users, but that itself is upfront work. Our recent moderation sprint cut down moderation effort a lot, but that was 3-4 weeks worth of work, and I think getting further gains would take at least similar time.
The other thing is it’s one thing to have a system under mild pressure vs intense pressure. With the later, I think we’d find we have to do lots of things to plug the gap.
There’s a question of whether huge influxes will happen involuntarily (somewhat my expectation), and in a world where we’ve built the tech to handle them, for sure also we should invite Rational Animations viewers.
It will be a while before we run an experiment, and when I’d like to start one, I’ll make another post and consult with you again.
When/if we do one, it’ll probably look like what @the gears to ascension proposed in their comment here: a pretty technical video that will likely get a smaller number of views than usual and filters for the kind of people we want on LessWrong. How I would advertise it could resemble the description of the market on Manifold linked in the post, but I’m going to run the details to you first.
LessWrong currently has about 2,000 logged-in users per day. And to 20-100 new users each day (comparing the wide range including peaks recently).
This provides important context. 20-100 new accounts per day is a lot. At the moment, Manifold predicts that as a result of a strong call to action and 1M views, Rational Animations would be able to bring 679 new expected users. That would probably look like getting 300-400 more users in the first couple weeks the video is out and an additional 300-400 in the following few months. That’s not a lot!
As a simplification, suppose the video gets 200k views during the first day. That would correspond to about 679⁄5 = 136 new expected users. Suppose on the second day we get 100k more views. That would be about 70 more users. Then suppose, simplifying, that the remaining 200k views are equally distributed over the remaining 12 days. That would correspond to merely 11 additional users per day.
Should LessWrong be providing intro material and answers? [...] So maybe. Maybe we should leaning into this as an opportunity even though it’ll take work of both not letting it affect the site in bad ways (moderation, etc). and also possibly preparing better material for a broader audience.
I would be happy to link such things if you produce them. For now, linking the AI Safety Fundamentals courses should achieve ~ the same results. Some of the readings can be found on LessWrong too so people may discover LW as a result too. That said, having something produced by LW probably improves the funnel.
To provide some insight: on the margin, more new users means more work for us. We process all first time posters/commenters manually, so there’s a linear factor there, and of new users, some require follow-up and then moderation action. So currently, there’s human cost in adding more people.
Duly noted. Another interesting datum would be to know the fraction of new users that become active posters and how long they take to do that.
First things first, I’m pro experiments so would be down to experiment with stuff in this area.
Beyond that, seems to depend on a couple of things:
The details of what inviting viewers to LW would look like.
What the LessWrong team thinks is the best use of our time.
The Details Matter
LessWrong currently has about 2,000 logged in users per day. And to 20-100 new users each day (comparing the wide range including peaks recently). If the numbers of viewers wouldn’t change that much, perhaps +10%, it wouldn’t be a big deal. On the other hand, if Rational Animations was wildly successful and driving several hundred people to create LW accounts a day (with flow through to posts/comments), that’d be a big deal. That would require work to handle.
Another big question is the level of sophistication of users. If you’re sending people to engage with more advanced stuff vs more intro content, that’s a big difference in how I expect them to relate to the site.
Should LessWrong be providing intro material and answers?
Even before you posted this, I’d been thinking about this question. I think it terms of knowledge and skill, we’re well positioned to provide a web resource of intro Alignment stuff. It’s not quite the core thing of “create a thriving intellectual community that can make progress on the important hard problems”, but might be a worthwhile enough opportunity that we should put time into it. I’d already been planning to do some work on our wiki system with an eye towards having more intro AI content for a broader audience.
So maybe. Maybe we should leaning into this as an opportunity even though it’ll take work of both not letting it affect the site in bad ways (moderation, etc). and also possibly preparing better material for a broader audience.
Moderation Costs
To provide some insight: on the margin, more new users means more work for us. We process all first time posters/commenters manually, so there’s a linear factor there, and of new users, some require follow-up and then moderation action. So currently, there’s human cost in adding more people.
It might be possible to build tech to reduce the marginal cost of new users, but that itself is upfront work. Our recent moderation sprint cut down moderation effort a lot, but that was 3-4 weeks worth of work, and I think getting further gains would take at least similar time.
The other thing is it’s one thing to have a system under mild pressure vs intense pressure. With the later, I think we’d find we have to do lots of things to plug the gap.
There’s a question of whether huge influxes will happen involuntarily (somewhat my expectation), and in a world where we’ve built the tech to handle them, for sure also we should invite Rational Animations viewers.
It will be a while before we run an experiment, and when I’d like to start one, I’ll make another post and consult with you again.
When/if we do one, it’ll probably look like what @the gears to ascension proposed in their comment here: a pretty technical video that will likely get a smaller number of views than usual and filters for the kind of people we want on LessWrong. How I would advertise it could resemble the description of the market on Manifold linked in the post, but I’m going to run the details to you first.
This provides important context. 20-100 new accounts per day is a lot. At the moment, Manifold predicts that as a result of a strong call to action and 1M views, Rational Animations would be able to bring 679 new expected users. That would probably look like getting 300-400 more users in the first couple weeks the video is out and an additional 300-400 in the following few months. That’s not a lot!
As a simplification, suppose the video gets 200k views during the first day. That would correspond to about 679⁄5 = 136 new expected users. Suppose on the second day we get 100k more views. That would be about 70 more users. Then suppose, simplifying, that the remaining 200k views are equally distributed over the remaining 12 days. That would correspond to merely 11 additional users per day.
I would be happy to link such things if you produce them. For now, linking the AI Safety Fundamentals courses should achieve ~ the same results. Some of the readings can be found on LessWrong too so people may discover LW as a result too. That said, having something produced by LW probably improves the funnel.
Duly noted. Another interesting datum would be to know the fraction of new users that become active posters and how long they take to do that.