In any case, it does seem LW curated posts and ACX posts both usually have neutral titles, especially given the occasionally contentious nature of their contents.
“Moldbug sold out” is definitely an attack on someone’s status. I still prefer it, because it makes a concrete claim about why. For instance, if the AI 2027 critique post title was “AI 2027′s Graphs Are Made Up And Unjustified” this would feel to me much better than something only about status like “AI 2027′s Timeline Forecasts Are Bad”.
For instance, if the AI 2027 critique post title was “AI 2027′s Graphs Are Made Up And Unjustified” this would feel to me much better than something only about status like “AI 2027′s Timeline Forecasts Are Bad”.
But then that wouldn’t be an accurate description of what titotal’s post is about.
“AI 2027′s authors’ arguments for superexponential growth curves are conceptually flawed, and their exponential model is neither exponential nor well-justified, and their graphs are made up and unjustified, and their projections don’t take into account many important variables, and benchmark+gaps is a worse model than the simplified one [for technical reasons], and these kinds of forecasts should be viewed with inherent skepticism for the following reasons” would be a proper summary of what is going on… but obviously it’s not suitable as a title.
I mean… the reason the AI 2027 critique isn’t titled “AI 2027′s Graphs Are Made Up And Unjustified” is obviously because the critique is about so much more than just some graphs on ACX and Twitter, right? That’s just one small part of the criticism, regardless of how much post-publication public discourse has focused on that one aspect.
The post is ultimately about why the timeline forecasts are (according to the author) bad, so it seems quite hard to title it something direct and concrete when it’s a compilation of many separate issues titotal has with AI 2027.
It’s difficult to determine what you would or wouldn’t call “directly writ[ing] the bottom line about the status of something in your title.”
titotal’s post was titled “A deep critique of AI 2027’s bad timeline models.” Is that more or less about the status of the bottom line than “Futarchy’s fundamental flaw” is? What about “Moldbug sold out” over on ACX?
In any case, it does seem LW curated posts and ACX posts both usually have neutral titles, especially given the occasionally contentious nature of their contents.
“Moldbug sold out” is definitely an attack on someone’s status. I still prefer it, because it makes a concrete claim about why. For instance, if the AI 2027 critique post title was “AI 2027′s Graphs Are Made Up And Unjustified” this would feel to me much better than something only about status like “AI 2027′s Timeline Forecasts Are Bad”.
Added: I searched through a bunch of ACX archives specifically for the word ‘bad’ in titles, I think both titles make a substantive claim about what is bad (Bad Definitions Of “Democracy” And “Accountability” Shade Into Totalitarianism and Perhaps It Is A Bad Thing That The World’s Leading AI Companies Cannot Control Their AIs, the latter of which is slightly sarcastic while making the object level claim that the AI companies cannot control their AIs).
Added2: It was easier to search the complete SSC history for ‘bad’. The examples are Bad Dreams, How Bad Are Things?, Asymmetric Weapons Gone Bad, and Response To Comments: The Tax Bill Is Still Very Bad, which was the sequel to The Tax Bill Compared To Other Very Expensive Things. The last one is the only one similar to what we’re discussing here, but in-context it is said in response to his commenters and as a sequel to a post which did a substantive thing, the title was not the primary thesis for the rest of the internet, which again seems different to me.
But then that wouldn’t be an accurate description of what titotal’s post is about.
“AI 2027′s authors’ arguments for superexponential growth curves are conceptually flawed, and their exponential model is neither exponential nor well-justified, and their graphs are made up and unjustified, and their projections don’t take into account many important variables, and benchmark+gaps is a worse model than the simplified one [for technical reasons], and these kinds of forecasts should be viewed with inherent skepticism for the following reasons” would be a proper summary of what is going on… but obviously it’s not suitable as a title.
I mean… the reason the AI 2027 critique isn’t titled “AI 2027′s Graphs Are Made Up And Unjustified” is obviously because the critique is about so much more than just some graphs on ACX and Twitter, right? That’s just one small part of the criticism, regardless of how much post-publication public discourse has focused on that one aspect.
The post is ultimately about why the timeline forecasts are (according to the author) bad, so it seems quite hard to title it something direct and concrete when it’s a compilation of many separate issues titotal has with AI 2027.