My perspective, Ray might have different thoughts on this:
“Scaling” in this post mostly refers to “scaling up from small (<20) closed-group interactions”. I have a lot more people than that that I want to actively talk to, and whose ideas I want to hear about in the context of existing LessWrong content.
Ultimately, a lot of the problems I care about seem hard, and it seems like we need at least a few dozen people working on them, maybe even a few hundred, if we want to solve them. It’s not obvious that there is super much to gain by scaling LessWrong to thousands of active commenters, but scaling it to at least a few dozen strikes me as necessary to solve a lot of the problems I care about (AI alignment, various open questions in individual rationality, various open questions in group rationality, etc.).
I care about the best people working on the problems I care about, and that requires an environment where a lot of people can try their hands at working on the problems, so that we can identify the ones that seem best and give them more resources and support. This requires a platform that can deal with new people coming in.
So scale it to...the size it already is? Maybe double that? I don’t think that requires any change. If you wanted a 10x user count increase, that probably would, but I don’t think those 10X potential users even exist. Unless and until round 3 of “Eliezer writes something that has no business getting a large audience into his preferred cause areas, but somehow works anyway” occurs.
I am also extremely skeptical that any discussion platform can do the third thing you mention. I don’t think any discussion platform that has ever existed both dealt with significant quantities of new people coming in well and was effective at filtering for effectiveness/quality. Those goals, in point of fact, seem directly opposed in most contexts; in order to judge people in any detail, the number to be judged must be kept small.
Are you sure you’re not building for scale because that’s the default thing you do with a web app made in the SF Bay Area?
Hmm, related question: Assuming this revival works, how long do you expect the site to be actively used before a 3.0 requiring a similar level of effort as this project becomes necessary? 5 years? 10?
5 years doesn’t strike me as insane. It seems that most online platforms require a makeover about once every 5 years, so yeah, if this goes well then launching LessWrong 3.0 in 2022-2023 seems quite reasonable. Though some platforms seem to last 10 years without makeover before they seriously decline, so maybe we can put it off until 2027-2028. I would be surprised if this website would be successful for more than 10 years without significant rework (the internet changes quickly, and there is a good chance we will all be browsing in VR by then, or social networks will have wholly consumed the whole internet, or some other change of the scale of the onset of social networks happens).
Yeah, it’s both important to me that the people I see doing the most valuable work on rationality and existential risk feel comfortable posting to the platform, and that we can continue replacing the people we will inevitably lose because of natural turnover with people of equal or better quality.
This has definitely not been the case over the previous 3 years of LessWrong, and so to fix that we will require some changes. My diagnosis of why that happened is partially that the nature of how people use the internet changed (with the onset of social networks and more competition due to better overall technology), partially because the people who were doing good work on the problems changed, and partially because we simply didn’t have a system that could productively fight against the forces of entropy for too long, and so a lot of the best people left.
I agree that it is hard to deal with large numbers of people joining at the same time, which is why I am indeed not super interested in discontinuous growth and am not pushing for anything in that direction. I do still think we are under the number of people who can productively talk to each other, and that at this point in time further-sustained, slow growth is valuable and net-positive.
Right now I think we are growing, though people have definitely been leaving over the last few years. I also think Eliezer has not had amazing experiences with the new LW, and there are some other people who showed up and would probably leave again if things don’t change in some way. On net I think we are strongly in the green, but I still think we are missing out on some really good and core people.
I’m thinking of it as “we’re growing, but on credit.” People are trying it out because they’ve heard enough interesting things to give it a go again, but it hasn’t hit something like genuine profitability.
Assuming this revival works, how long do you expect the site to be actively used before a 3.0 requiring a similar level of effort as this project becomes necessary? 5 years? 10? (My prediction is 5 years.)
There’s a lot of writing about how to make predictions for the future on LessWrong and this is a poor one. Good predictions have probabilities attached to them.
Is your concern that it’s not clear whether PDV’s estimate is a mean, median, or mode? “Median and mode” seems like a reasonable guess (though have to be careful when defining mode etc.)
Being ambiguous about your prediction leaves wiggle room, but that’s typical for english sentences; giving an estimate without saying what exactly it means is still less ambiguous than the default.
The question is a bit like “When will you stop beating your wife?”. It assumes that some time in the future there will be a need to invest resources in a LW 3.0.
That’s a bad way to think about the future. It’s much better to think about the future by thinking that different event might happen and those have probabilities attached to them.
I mean, I mostly agree with PDVs prediction here. Very few things on the internet survive for more than 5 years, and so yeah, I think it’s likely you will need a group of at least two or three full-time people to work on it to ensure that it stays active and healthy and up-to-date with the way people use the internet in 5 years.
Why do you think that LessWrong can or should scale?
My perspective, Ray might have different thoughts on this:
“Scaling” in this post mostly refers to “scaling up from small (<20) closed-group interactions”. I have a lot more people than that that I want to actively talk to, and whose ideas I want to hear about in the context of existing LessWrong content.
Ultimately, a lot of the problems I care about seem hard, and it seems like we need at least a few dozen people working on them, maybe even a few hundred, if we want to solve them. It’s not obvious that there is super much to gain by scaling LessWrong to thousands of active commenters, but scaling it to at least a few dozen strikes me as necessary to solve a lot of the problems I care about (AI alignment, various open questions in individual rationality, various open questions in group rationality, etc.).
I care about the best people working on the problems I care about, and that requires an environment where a lot of people can try their hands at working on the problems, so that we can identify the ones that seem best and give them more resources and support. This requires a platform that can deal with new people coming in.
So scale it to...the size it already is? Maybe double that? I don’t think that requires any change. If you wanted a 10x user count increase, that probably would, but I don’t think those 10X potential users even exist. Unless and until round 3 of “Eliezer writes something that has no business getting a large audience into his preferred cause areas, but somehow works anyway” occurs.
I am also extremely skeptical that any discussion platform can do the third thing you mention. I don’t think any discussion platform that has ever existed both dealt with significant quantities of new people coming in well and was effective at filtering for effectiveness/quality. Those goals, in point of fact, seem directly opposed in most contexts; in order to judge people in any detail, the number to be judged must be kept small.
Are you sure you’re not building for scale because that’s the default thing you do with a web app made in the SF Bay Area?
Hmm, related question: Assuming this revival works, how long do you expect the site to be actively used before a 3.0 requiring a similar level of effort as this project becomes necessary? 5 years? 10?
(My prediction is 5 years.)
5 years doesn’t strike me as insane. It seems that most online platforms require a makeover about once every 5 years, so yeah, if this goes well then launching LessWrong 3.0 in 2022-2023 seems quite reasonable. Though some platforms seem to last 10 years without makeover before they seriously decline, so maybe we can put it off until 2027-2028. I would be surprised if this website would be successful for more than 10 years without significant rework (the internet changes quickly, and there is a good chance we will all be browsing in VR by then, or social networks will have wholly consumed the whole internet, or some other change of the scale of the onset of social networks happens).
If people are leaving as we speak, then scaling it to the size it already is may indeed require change.
Yeah, it’s both important to me that the people I see doing the most valuable work on rationality and existential risk feel comfortable posting to the platform, and that we can continue replacing the people we will inevitably lose because of natural turnover with people of equal or better quality.
This has definitely not been the case over the previous 3 years of LessWrong, and so to fix that we will require some changes. My diagnosis of why that happened is partially that the nature of how people use the internet changed (with the onset of social networks and more competition due to better overall technology), partially because the people who were doing good work on the problems changed, and partially because we simply didn’t have a system that could productively fight against the forces of entropy for too long, and so a lot of the best people left.
I agree that it is hard to deal with large numbers of people joining at the same time, which is why I am indeed not super interested in discontinuous growth and am not pushing for anything in that direction. I do still think we are under the number of people who can productively talk to each other, and that at this point in time further-sustained, slow growth is valuable and net-positive.
Do you think that people are leaving at more than a reasonable rate of natural attrition? If so, why?
Right now I think we are growing, though people have definitely been leaving over the last few years. I also think Eliezer has not had amazing experiences with the new LW, and there are some other people who showed up and would probably leave again if things don’t change in some way. On net I think we are strongly in the green, but I still think we are missing out on some really good and core people.
I’m thinking of it as “we’re growing, but on credit.” People are trying it out because they’ve heard enough interesting things to give it a go again, but it hasn’t hit something like genuine profitability.
There’s a lot of writing about how to make predictions for the future on LessWrong and this is a poor one. Good predictions have probabilities attached to them.
Is your concern that it’s not clear whether PDV’s estimate is a mean, median, or mode? “Median and mode” seems like a reasonable guess (though have to be careful when defining mode etc.)
Being ambiguous about your prediction leaves wiggle room, but that’s typical for english sentences; giving an estimate without saying what exactly it means is still less ambiguous than the default.
The question is a bit like “When will you stop beating your wife?”. It assumes that some time in the future there will be a need to invest resources in a LW 3.0.
That’s a bad way to think about the future. It’s much better to think about the future by thinking that different event might happen and those have probabilities attached to them.
I mean, I mostly agree with PDVs prediction here. Very few things on the internet survive for more than 5 years, and so yeah, I think it’s likely you will need a group of at least two or three full-time people to work on it to ensure that it stays active and healthy and up-to-date with the way people use the internet in 5 years.