But the clicks! The smaller the baserate, the less impressive the absolute change, but the bigger the percent change.
Percents get the clicks!
But the clicks! The smaller the baserate, the less impressive the absolute change, but the bigger the percent change.
Percents get the clicks!
The daily numbers are on X: x.com/BaseRateAI
The daily numbers are on X: x.com/BaseRateAI
I have 5 slots available, DM me if you’re interested! And as always, I appreciate any ♥️/ ↪️!
I have 5 slots available, DM me if you’re interested! And as always, I appreciate any ♥️/ ↪️!
& species often get outcompeted by being reconnected with near-speciated rivals, I tried to get a baserate for this (endogenous vs. exogenous) but failed
& species often get outcompeted by being reconnected with near-speciated rivals, I tried to get a baserate for this (endogenous vs. exogenous) but failed
These numbers mostly agree with other estimates. nobody would need to be a mind-reader to detect a serious trend like "a major fraction of the millions of working SWEs are experiencing mass-psychosis." I know *zero*, which sounds like the baserate
These numbers mostly agree with other estimates. nobody would need to be a mind-reader to detect a serious trend like "a major fraction of the millions of working SWEs are experiencing mass-psychosis." I know *zero*, which sounds like the baserate
www.rivm.nl/publicat...
2/10
www.rivm.nl/publicat...
2/10
& @noosphere89.bsky.social ¹):
LLMs often think they're in an eval when they aren't, probabilities reported: 15%, 15%, 2%, 3%, 25%–40%, 15%–25%, 60%–75%, 10–15%, 5-10%, 70-80%, 55-70%
This is way higher than what a baserate should say, which is ~10⁻⁴–10⁻⁷
& @noosphere89.bsky.social ¹):
LLMs often think they're in an eval when they aren't, probabilities reported: 15%, 15%, 2%, 3%, 25%–40%, 15%–25%, 60%–75%, 10–15%, 5-10%, 70-80%, 55-70%
This is way higher than what a baserate should say, which is ~10⁻⁴–10⁻⁷
But it's still likely a pretty inefficient one (in part bc of that baserate issue).
But it's still likely a pretty inefficient one (in part bc of that baserate issue).
My usual rebuttal to “biggest percent change!” is “yeah, but what abt the absolute change?”
Which here? Also the biggest of all time.
My usual rebuttal to “biggest percent change!” is “yeah, but what abt the absolute change?”
Which here? Also the biggest of all time.
Jonathan Haidt support? ✅
Article later admits most kids are actually taking types of non-home classes, just in a different form, bc admit officers don’t trust parent grades? ✅
Clear selection bias of wealthy parents w big tutoring and college counseling budgets? ✅
Jonathan Haidt support? ✅
Article later admits most kids are actually taking types of non-home classes, just in a different form, bc admit officers don’t trust parent grades? ✅
Clear selection bias of wealthy parents w big tutoring and college counseling budgets? ✅
But the absolute decline *just since 2022* has been the biggest in modern history—even compared to the 2000s, which has a comparable baserate.
And crickets on that too.
So it’s not a rates-vs-absolutes issue—it’s a deeper problem.
But the absolute decline *just since 2022* has been the biggest in modern history—even compared to the 2000s, which has a comparable baserate.
And crickets on that too.
So it’s not a rates-vs-absolutes issue—it’s a deeper problem.
The baserate fallacy is real and omnipresent.
(The post I’m subtweeting isn’t worth flagging, but any time I see that line I feel the need to raise the baserate fallacy alarm.)
The baserate fallacy is real and omnipresent.
(The post I’m subtweeting isn’t worth flagging, but any time I see that line I feel the need to raise the baserate fallacy alarm.)
I swear to God, percentages and the baserate problem are one of the single greatest threats to sound policy, and I’m not really being hyperbolic.
MEMPHIS POLICE: “Those crimes are actually down.”
🤷🏼♀️ 🚂 (h/t @NoApologies901)
I swear to God, percentages and the baserate problem are one of the single greatest threats to sound policy, and I’m not really being hyperbolic.
The way we incarcerate women (esp what it says abt who gets the self-defense benefit of the doubt and who doesn’t) is terrible.
But the narrative around it has some baserate issues too.
The way we incarcerate women (esp what it says abt who gets the self-defense benefit of the doubt and who doesn’t) is terrible.
But the narrative around it has some baserate issues too.
When this baserate issue doesn't hold--when a huge percent change = huge absolute change--it's something huge and catastrophic we don't need data to know (see: Covid).
Also, "assault" can be defined REALLY broadly.
When this baserate issue doesn't hold--when a huge percent change = huge absolute change--it's something huge and catastrophic we don't need data to know (see: Covid).
Also, "assault" can be defined REALLY broadly.
This is one of the only polls in recent memory that I can think of where the results aren’t just “% of respondents who are Dem [or GOP] +/- a small error.”
For once, “twice as many GOPers!” is a baserate issue, not a “The City and The City” issue.
And the partisan splits (or relative lack there of) manage to be even more impressive
This is one of the only polls in recent memory that I can think of where the results aren’t just “% of respondents who are Dem [or GOP] +/- a small error.”
For once, “twice as many GOPers!” is a baserate issue, not a “The City and The City” issue.
Also, desert = county w no local news. Lots of “deserts” are in TX. Ever seen a county-level TX map? TX, home to nearly 10% of all *US* counties?
Also, desert = county w no local news. Lots of “deserts” are in TX. Ever seen a county-level TX map? TX, home to nearly 10% of all *US* counties?