This is going to sound controversial here (people are probably going to dislike this but I’m genuinely raising this as a concern) but is the Robert Miles $60,000 grant attached to any requirements? I like his content but it seems to me you could find someone with a similar talent level (explaining fairly basic concepts) who could produce many more videos. I’m not well versed in YouTube but four/five videos in the last year doesn’t seem substantial. If the $60,000 was instead offered as a one-year job, I think you could find many talented individuals who could produce much more content.
I understand that he’s doing other non-directly YouTube related things but if you include support in other forms (Patreon), the output seems pretty low relative to the investment.
Again I should emphasise I’m uncertain about my criticism here and personally have enjoyed watching his videos on occasion.
I think one of the things Rob has that is very hard to replace is his audience. Overall I continue to be shocked by the level of engagement Rob Miles’ youtube videos get. Averaging over 100k views per video! I mostly disbelieve that it would be plausible to hire someone that can (a) understand technical AI alignment well, and (b) reliably create youtube videos that get over 100k views, for less than something like an order of magnitude higher cost.
I am mostly confused about how Rob gets 100k+ views on each video. My mainline hypothesis is that Rob has successfully built his own audience through his years of videos including on places like Computerphile, and that they have followed him to his own channel.
Building an audience like this takes many years and often does not pay off. Once you have a massive audience that cares about the kind of content you produce, this is very quickly not replaceable, and I expect to find someone other than Rob to do this, it would either take the person 3-10 years to build this size of audience, or require paying a successful youtube content creator to change the videos that they are making substantially, in a way that risks losing their audience, and thus require a lot of money to cover the risk (I’m imagining $300k–$1mil per year for the first few years).
Another person to think of here is Tim Urban, who writes Wait But Why. That blog has I think produced zero major writeups in the last year, but he has a massive audience who knows him and is very excited to read his content in detail, which is valuable and not easily replaceable. If it were possible to pay Tim Urban to write a piece on a technical topic of your choice, this would be exceedingly widely-read in detail, and would be worth a lot of money even if he didn’t publish anything for a whole year.
All good points Jonas, Ben W, Ben P, and Stefan. Was uncertain at the beginning but am pretty convinced now. Also, side-note, very happy about the nature of all of the comments, in that they understood my POV and engaged with them in a polite manner.
By the way, I also was surprised by Rob only making 4 videos in the last year. But I actually now think Rob is producing a fairly standard number of high-quality videos annually.
The first reason is that (as Jonas points out upthread) he also did three for Computerphile, which brings his total to 7.
The second reason is that I looked into a bunch of top YouTube individual explainers, and I found that they produce a similar number of highly-produced videos annually. Here’s a few:
3 Blue 1 Brown has 10 highly produced videos in the last year (1, 2, 3, 4, 5, 6, 7, 8, 9, 10). He has other videos, which include a vide of Grant talking a walk, a short footnote video to one of the main ones, 10 lockdown livestream videos, and a video turning someone’s covid blogpost into a video. For highly produced videos, he’s averaging just under 1/month.
CGP Grey has 10 highly produced videos in the last year (1, 2, 3, 4, 5, 6, 7, 8, 9, 10). He has other videos, which include a video of CGP Grey talking a walk, a few videos of him exploring a thing like a spreadsheet or an old building, and one or two commentaries on other videos of his.
Vi Hart in her peak made 19 videos in one year (her first year, 9 years ago) all of which I think were of a similar quality level to each other.
Veritasium has 14 highlighy produced videos in the last year, plus one short video of the creator monologuing after their visit to NASA.
CGP Grey, 3Blue 1Brown and Veritasium I believe are working on their videos full time, so I think around 10 main videos plus assorted extra pieces is within standard range for highly successful explainers on YouTube. I think this suggests potentially Rob could make more videos to fill out the space between the videos on his channel, like Q&A livestreams and other small curiosities that he notices, and could plausibly be more productive a year in terms of making a couple more of the main, highly-produced videos.
But I know he does a fair bit of other work outside of his main channel, and also he is in some respects doing a harder task than some of the above, of explaining ideas from a new research field, and one with a lot of ethical concerns around the work, not just issues of how to explain things well, which I expect increases the amount of work that goes into the videos.
I think it’s possible that last year was just unusually slow for people (possibly pandemic-related?)
I looked at 3B1B (the only Youtube explainer series I’m familiar with) and since 2015 Grant has produced ~100 high quality videos, which is closer to ~20 videos/year than ~10/year.
I’m not familiar with the others.
and could plausibly be ~20% more productive in a year in terms of the main, highly-produced videos
I feel like this is low-balling potential year-to-year variation in productivity. My inside view is that 50-100% increases in productivity is plausible.
Yeah, I agree about how much variance in productivity is available, your numbers seem more reasonable. I’d actually edited it by the time you wrote your comment.
Also agree last year was probably unusually slow all round. I expect the comparison is still comparing like-with-like.
In addition to four videos on his own channel, Robert Miles also published threevideoson Computerphile during the last 12 months. He also publishes the Alignment Newsletter podcast. So there’s at least some additional output. There’s probably more I don’t know of.
you could find someone with a similar talent level (explaining fairly basic concepts)
I personally actually think this would be very difficult. Robert Miles’ content seems to have been received positively by the AI safety community, but science communications in general is notoriously difficult, and I’d expect most YouTubers to routinely distort and oversimplify important concepts, such that I’d worry that such content would do more harm than good. In contrast, Robert Miles seems sufficiently nuanced.
you could find someone with a similar talent level … who could produce many more videos
It seems that the Long-Term Future Fund isn’t actively searching for people to do specific tasks, if I understand the post correctly. Instead, it’s reviewing applications that come to them. (It’s more labour-intensive to do an active search.) That means that it can be warranted to fund an applicant even if it’s possible that there could be better candidates for the same task somewhere out there. (Minor edits.)
Thanks for the understanding responses Jonas and Linch. Again, I should clarify, I don’t know where I stand here but I’m still not entirely convinced.
So, we have four videos in the last year on his channel, plus three videos on Computerphile, giving seven videos. If I remember correctly, The Alignment Newsletter podcast is just reading Shah’s newsletter, which may be useful but I don’t think that requires a lot of effort.
I should reiterate that I think what Miles does is not easy. I may also be severely underestimating the time it takes to make a YouTube video!
It might be more relevant to consider the output: 500,000 views (or ~80,000 hours of watch time). Given that the median video gets 89 views, it might be hard for other creators to match the output, even if they could produce more videos per se.
Thanks for pointing that out. Will refrain from doing so in the future. What I was trying to make clear was that I didn’t want my comment to be seen as a personal attack on an individual. I was uneasy about making the comment on a public platform when I don’t know all the details nor know much about the subject matter.
FWIW, I think that the qualification was very appropriate and I didn’t see the author as intending to start a “bravery debate”. Instead, the purpose appears to have been to emphasize that the concerns were raised in good faith and with limited information. Clarifications of this sort seem very relevant and useful, and quite unlike the phenomenon described in Scott’s post.
I want to add that Scott isn’t describing a disingenuous argumentative tactic, he’s saying that the topic causes dialogue to get derailed very quickly. Analogous to the rule that bringing in a comparison to Nazis always derails internet discussion, making claims about whether the position one is advocating is the underdog or the mainstream also derails internet discussion.
Following up, and sorry for continuing to critique after you already politely made an edit, but doesn’t that change your opinion of the object level thing, which is indeed the phenomenon Scott’s talking about? It’s great to send signals of cooperativeness and genuineness, and I appreciate So-Low Growth’s effort to do so, but adding in talk of how the concern is controversial is the standard example of opening a bravery debate.
The application of Scott’s post here would be to separate clarification of intent and bravery talk – in this situation, separating “I don’t intend any personal attack on this individual” from “My position is unpopular”. Again, the intention is not in question, it’s the topic, and that’s the phenomenon Scott’s discussing in his post.
I agree that the sentence Linch quoted sounds like a “bravery debate” opening, but that’s not how I perceive it in the broader context. I don’t think the author is presenting himself/herself as an underdog, intentionally or otherwise. Rather, they are making that remark as part of their overall attempt to indicate that they are aware that they are raising a sensitive issue and that they are doing so in a collaborative spirit and with admittedly limited information. This strikes me as importantly different from the prototypical bravery debate, where the primary effect is not to foster an atmosphere of open dialogue but to gain sympathy for a position.
I am tentatively in agreement with you that “clarification of intent” can be done without “bravery talk”, by which I understand any mention that the view one is advancing is unpopular. But I also think that such talk doesn’t always communicate that one is the underdog, and is therefore not inherently problematic. So, yes, the OP could have avoided that kind of language altogether, but given the broader context, I don’t think the use of that language did any harm.
(I’m maybe 80% confident in what I say above, so if you disagree, feel free to push me.)
I read the top comment again after reading this comment by you, and I think I understand the original intent better now. I was mostly confused on initial reading, and while I thought SLG’s comment was otherwise good and I had a high prior on the intent being very cooperative, I couldn’t figure out what the first line meant other than “I expect I’m the underdog here”. I now read it as saying “I really don’t want to cause conflict needlessly, but I do care about discussing this topic,” which seems pretty positive to me. I am pretty pro SLG writing more comments like this in future when it seems to them like an important mistake is likely being made :)
This makes a lot of sense to me Pablo. You highlighted what I was trying to explain when I was making the comment, that: 1) I was uncertain 2) I didn’t want to attack someone. I must admit, my choice of words was rather poor and could come across as “bravery talk”, although that was not what I intended.
To be clear, I think your overall comment added to the discussion more than it detracts, and I really appreciate you making it. I definitely did not interpret your claims as an attack, nor did I think it’s a particularly egregious example of a bravery framing. One reason I chose to comment here is because I interpreted (correctly, it appears!) you as someone who’d be receptive to such feedback, whereas if somebody started a bravery debate with a clearer “me against the immoral idiots in EA” framing I’d probably be much more inclined to just ignore and move on.
It’s possible my bar for criticism is too low. In particular, I don’t think I’ve fully modeled meta-level considerations like:
1) That by only choosing to criticize mild rather than egregious cases, I’m creating bad incentives.
2) You appear to be a new commenter, and by criticizing newcomers to the EA Forum I risk making the EA Forum less appealing.
3) That my comment may spawn a long discussion.
Nonetheless I think I mostly stand by my original comment.
Yeah that makes a lot of sense. I think the rest of your comment is fine without that initial disclaimer, especially with your caveat in the last sentence! :)
I also notice myself being confused about the output here. I suspect the difficulty of being good at Youtube outreach while fully understanding technical AI safety concepts is a higher bar than you’re claiming, but I also intuitively would be surprised if it takes an average of 2+ months to produce a video (though perhaps he spends a lot of time on other activities?
This quote
for example, he’s already helping existing organizations produce videos about their ideas
To state a point in the neighborhood of what Stefan, Ben P, and Ben W have said, I think it’s important for LTTF to evaluate the counterfactual where they don’t fund something, rather than the counterfactual where the project has more reasonable characteristics.
That is, we might prefer a project be more productive, more legible or more organized, but unless that makes it worse than the marginal funding opportunity, it should be funded (where one way a project could be bad is by displacing more reasonable projects that would otherwise fill a gap).
This is going to sound controversial here (people are probably going to dislike this but I’m genuinely raising this as a concern) but is the Robert Miles $60,000 grant attached to any requirements? I like his content but it seems to me you could find someone with a similar talent level (explaining fairly basic concepts) who could produce many more videos. I’m not well versed in YouTube but four/five videos in the last year doesn’t seem substantial. If the $60,000 was instead offered as a one-year job, I think you could find many talented individuals who could produce much more content.
I understand that he’s doing other non-directly YouTube related things but if you include support in other forms (Patreon), the output seems pretty low relative to the investment.
Again I should emphasise I’m uncertain about my criticism here and personally have enjoyed watching his videos on occasion.
I think one of the things Rob has that is very hard to replace is his audience. Overall I continue to be shocked by the level of engagement Rob Miles’ youtube videos get. Averaging over 100k views per video! I mostly disbelieve that it would be plausible to hire someone that can (a) understand technical AI alignment well, and (b) reliably create youtube videos that get over 100k views, for less than something like an order of magnitude higher cost.
I am mostly confused about how Rob gets 100k+ views on each video. My mainline hypothesis is that Rob has successfully built his own audience through his years of videos including on places like Computerphile, and that they have followed him to his own channel.
Building an audience like this takes many years and often does not pay off. Once you have a massive audience that cares about the kind of content you produce, this is very quickly not replaceable, and I expect to find someone other than Rob to do this, it would either take the person 3-10 years to build this size of audience, or require paying a successful youtube content creator to change the videos that they are making substantially, in a way that risks losing their audience, and thus require a lot of money to cover the risk (I’m imagining $300k–$1mil per year for the first few years).
Another person to think of here is Tim Urban, who writes Wait But Why. That blog has I think produced zero major writeups in the last year, but he has a massive audience who knows him and is very excited to read his content in detail, which is valuable and not easily replaceable. If it were possible to pay Tim Urban to write a piece on a technical topic of your choice, this would be exceedingly widely-read in detail, and would be worth a lot of money even if he didn’t publish anything for a whole year.
All good points Jonas, Ben W, Ben P, and Stefan. Was uncertain at the beginning but am pretty convinced now. Also, side-note, very happy about the nature of all of the comments, in that they understood my POV and engaged with them in a polite manner.
By the way, I also was surprised by Rob only making 4 videos in the last year. But I actually now think Rob is producing a fairly standard number of high-quality videos annually.
The first reason is that (as Jonas points out upthread) he also did three for Computerphile, which brings his total to 7.
The second reason is that I looked into a bunch of top YouTube individual explainers, and I found that they produce a similar number of highly-produced videos annually. Here’s a few:
3 Blue 1 Brown has 10 highly produced videos in the last year (1, 2, 3, 4, 5, 6, 7, 8, 9, 10). He has other videos, which include a vide of Grant talking a walk, a short footnote video to one of the main ones, 10 lockdown livestream videos, and a video turning someone’s covid blogpost into a video. For highly produced videos, he’s averaging just under 1/month.
CGP Grey has 10 highly produced videos in the last year (1, 2, 3, 4, 5, 6, 7, 8, 9, 10). He has other videos, which include a video of CGP Grey talking a walk, a few videos of him exploring a thing like a spreadsheet or an old building, and one or two commentaries on other videos of his.
Vi Hart in her peak made 19 videos in one year (her first year, 9 years ago) all of which I think were of a similar quality level to each other.
Veritasium has 14 highlighy produced videos in the last year, plus one short video of the creator monologuing after their visit to NASA.
CGP Grey, 3Blue 1Brown and Veritasium I believe are working on their videos full time, so I think around 10 main videos plus assorted extra pieces is within standard range for highly successful explainers on YouTube. I think this suggests potentially Rob could make more videos to fill out the space between the videos on his channel, like Q&A livestreams and other small curiosities that he notices, and could plausibly be more productive a year in terms of making a couple more of the main, highly-produced videos.
But I know he does a fair bit of other work outside of his main channel, and also he is in some respects doing a harder task than some of the above, of explaining ideas from a new research field, and one with a lot of ethical concerns around the work, not just issues of how to explain things well, which I expect increases the amount of work that goes into the videos.
I think it’s possible that last year was just unusually slow for people (possibly pandemic-related?)
I looked at 3B1B (the only Youtube explainer series I’m familiar with) and since 2015 Grant has produced ~100 high quality videos, which is closer to ~20 videos/year than ~10/year.
I’m not familiar with the others.
I feel like this is low-balling potential year-to-year variation in productivity. My inside view is that 50-100% increases in productivity is plausible.
Yeah, I agree about how much variance in productivity is available, your numbers seem more reasonable. I’d actually edited it by the time you wrote your comment.
Also agree last year was probably unusually slow all round. I expect the comparison is still comparing like-with-like.
:) Appreciated the conversation! It also gave me an opportunity to clarify my own thoughts about success on YouTube and related things.
Thanks for the critique!
In addition to four videos on his own channel, Robert Miles also published three videos on Computerphile during the last 12 months. He also publishes the Alignment Newsletter podcast. So there’s at least some additional output. There’s probably more I don’t know of.
I personally actually think this would be very difficult. Robert Miles’ content seems to have been received positively by the AI safety community, but science communications in general is notoriously difficult, and I’d expect most YouTubers to routinely distort and oversimplify important concepts, such that I’d worry that such content would do more harm than good. In contrast, Robert Miles seems sufficiently nuanced.
(Disclosure: I work at EA Funds.)
Yes. Also, regarding this issue:
It seems that the Long-Term Future Fund isn’t actively searching for people to do specific tasks, if I understand the post correctly. Instead, it’s reviewing applications that come to them. (It’s more labour-intensive to do an active search.) That means that it can be warranted to fund an applicant even if it’s possible that there could be better candidates for the same task somewhere out there. (Minor edits.)
Thanks for the understanding responses Jonas and Linch. Again, I should clarify, I don’t know where I stand here but I’m still not entirely convinced.
So, we have four videos in the last year on his channel, plus three videos on Computerphile, giving seven videos. If I remember correctly, The Alignment Newsletter podcast is just reading Shah’s newsletter, which may be useful but I don’t think that requires a lot of effort.
I should reiterate that I think what Miles does is not easy. I may also be severely underestimating the time it takes to make a YouTube video!
It might be more relevant to consider the output: 500,000 views (or ~80,000 hours of watch time). Given that the median video gets 89 views, it might be hard for other creators to match the output, even if they could produce more videos per se.
Meta: Small nitpick, but I would prefer if we reduce framings like
See Scott Alexander on Against Bravery Debates.
Thanks for pointing that out. Will refrain from doing so in the future. What I was trying to make clear was that I didn’t want my comment to be seen as a personal attack on an individual. I was uneasy about making the comment on a public platform when I don’t know all the details nor know much about the subject matter.
FWIW, I think that the qualification was very appropriate and I didn’t see the author as intending to start a “bravery debate”. Instead, the purpose appears to have been to emphasize that the concerns were raised in good faith and with limited information. Clarifications of this sort seem very relevant and useful, and quite unlike the phenomenon described in Scott’s post.
I want to add that Scott isn’t describing a disingenuous argumentative tactic, he’s saying that the topic causes dialogue to get derailed very quickly. Analogous to the rule that bringing in a comparison to Nazis always derails internet discussion, making claims about whether the position one is advocating is the underdog or the mainstream also derails internet discussion.
Thanks, you are right. I have amended the last sentence of my comment.
Thx!
Following up, and sorry for continuing to critique after you already politely made an edit, but doesn’t that change your opinion of the object level thing, which is indeed the phenomenon Scott’s talking about? It’s great to send signals of cooperativeness and genuineness, and I appreciate So-Low Growth’s effort to do so, but adding in talk of how the concern is controversial is the standard example of opening a bravery debate.
The application of Scott’s post here would be to separate clarification of intent and bravery talk – in this situation, separating “I don’t intend any personal attack on this individual” from “My position is unpopular”. Again, the intention is not in question, it’s the topic, and that’s the phenomenon Scott’s discussing in his post.
I agree that the sentence Linch quoted sounds like a “bravery debate” opening, but that’s not how I perceive it in the broader context. I don’t think the author is presenting himself/herself as an underdog, intentionally or otherwise. Rather, they are making that remark as part of their overall attempt to indicate that they are aware that they are raising a sensitive issue and that they are doing so in a collaborative spirit and with admittedly limited information. This strikes me as importantly different from the prototypical bravery debate, where the primary effect is not to foster an atmosphere of open dialogue but to gain sympathy for a position.
I am tentatively in agreement with you that “clarification of intent” can be done without “bravery talk”, by which I understand any mention that the view one is advancing is unpopular. But I also think that such talk doesn’t always communicate that one is the underdog, and is therefore not inherently problematic. So, yes, the OP could have avoided that kind of language altogether, but given the broader context, I don’t think the use of that language did any harm.
(I’m maybe 80% confident in what I say above, so if you disagree, feel free to push me.)
I read the top comment again after reading this comment by you, and I think I understand the original intent better now. I was mostly confused on initial reading, and while I thought SLG’s comment was otherwise good and I had a high prior on the intent being very cooperative, I couldn’t figure out what the first line meant other than “I expect I’m the underdog here”. I now read it as saying “I really don’t want to cause conflict needlessly, but I do care about discussing this topic,” which seems pretty positive to me. I am pretty pro SLG writing more comments like this in future when it seems to them like an important mistake is likely being made :)
This makes a lot of sense to me Pablo. You highlighted what I was trying to explain when I was making the comment, that: 1) I was uncertain 2) I didn’t want to attack someone. I must admit, my choice of words was rather poor and could come across as “bravery talk”, although that was not what I intended.
To be clear, I think your overall comment added to the discussion more than it detracts, and I really appreciate you making it. I definitely did not interpret your claims as an attack, nor did I think it’s a particularly egregious example of a bravery framing. One reason I chose to comment here is because I interpreted (correctly, it appears!) you as someone who’d be receptive to such feedback, whereas if somebody started a bravery debate with a clearer “me against the immoral idiots in EA” framing I’d probably be much more inclined to just ignore and move on.
It’s possible my bar for criticism is too low. In particular, I don’t think I’ve fully modeled meta-level considerations like:
1) That by only choosing to criticize mild rather than egregious cases, I’m creating bad incentives.
2) You appear to be a new commenter, and by criticizing newcomers to the EA Forum I risk making the EA Forum less appealing.
3) That my comment may spawn a long discussion.
Nonetheless I think I mostly stand by my original comment.
Yeah that makes a lot of sense. I think the rest of your comment is fine without that initial disclaimer, especially with your caveat in the last sentence! :)
I also notice myself being confused about the output here. I suspect the difficulty of being good at Youtube outreach while fully understanding technical AI safety concepts is a higher bar than you’re claiming, but I also intuitively would be surprised if it takes an average of 2+ months to produce a video (though perhaps he spends a lot of time on other activities?
This quote
alludes to this.
To state a point in the neighborhood of what Stefan, Ben P, and Ben W have said, I think it’s important for LTTF to evaluate the counterfactual where they don’t fund something, rather than the counterfactual where the project has more reasonable characteristics.
That is, we might prefer a project be more productive, more legible or more organized, but unless that makes it worse than the marginal funding opportunity, it should be funded (where one way a project could be bad is by displacing more reasonable projects that would otherwise fill a gap).