Favorite part is where she spends half of the video talking about how stupid she was for allowing others tell her what to think,
But then one day it all fortunately changed when finally
…some dude told her what to think instead.
Literally the meme.
🐻🐻🐻
But then one day it all fortunately changed when finally
…some dude told her what to think instead.
Literally the meme.
🐻🐻🐻
💯7
Please open Telegram to view this post
VIEW IN TELEGRAM
No twitter stats guy more fails to do proper analysis
— that could properly detecting midwit curves that slope up at the extremes,
Instead of concluding that the curve slopes the OPPOSITE direction,
than midwit Crémieux over here.
Been meaning to point this out,
— Has anyone ever explicitly realized that the classic linear regression curves essentially ALWAYS get the slope exactly backward,
whenever there’s a midwit curve?
And as we know, midwit curves have habit of surprisingly showing up virtually everywhere there’s some controversial.
About time to rise up against the midwits, and their intentionally-flawed analysis procedures.
— that could properly detecting midwit curves that slope up at the extremes,
Instead of concluding that the curve slopes the OPPOSITE direction,
than midwit Crémieux over here.
Been meaning to point this out,
— Has anyone ever explicitly realized that the classic linear regression curves essentially ALWAYS get the slope exactly backward,
whenever there’s a midwit curve?
And as we know, midwit curves have habit of surprisingly showing up virtually everywhere there’s some controversial.
About time to rise up against the midwits, and their intentionally-flawed analysis procedures.
You know what — autist time.
Got here the original data from that famous 2021 vaccine acceptance midwit curve study.
Let’s get ChatGPT graph it, and I’ll show you exactly what I mean
— Perhaps the biggest type of lie used by the midwit science scammers today,
And AFAIK, a lie that no one has ever explicitly pointed out, yet.
So simple, so extreme, so unrecognized.
Vaccine acceptance study
Got here the original data from that famous 2021 vaccine acceptance midwit curve study.
Let’s get ChatGPT graph it, and I’ll show you exactly what I mean
— Perhaps the biggest type of lie used by the midwit science scammers today,
And AFAIK, a lie that no one has ever explicitly pointed out, yet.
So simple, so extreme, so unrecognized.
Vaccine acceptance study
DoomPosting
You know what — autist time. Got here the original data from that famous 2021 vaccine acceptance midwit curve study. Let’s get ChatGPT graph it, and I’ll show you exactly what I mean — Perhaps the biggest type of lie used by the midwit science scammers…
Continued: The “Smarter people more likely to believe X lie”
So, then we paste in the exact data from the study,
and then do the 2 types of simple analysis we’re wondering about here:
(1) Regular linear regression — I.e. what today’s social “scientists” use to answer questions of the form:
“The smarter you are, the more likely you agree/disagree with X.”
(2) Midwit-aware linear regression — A slightly modified linear regression you’d need to use, were life full of midwit curves, to answer questions of the form:
“Once you cross a sufficiently smart threshold, where the getting smarter won’t make you flip your answer again anymore — then the more you continue toward unboundedly-high smartness, the more you agree/disagree with X.”
Basically, in both cases, we’re looking for:
“Smarter people believe X”.
So then we have ChatGPT run this analysis on the real data from the study
…and…
WTF IS THIS.
LOOK HOW OPPOSITE THOSE LINES ARE.
EVEN WORSE THAN I IMAGINED.
THEY’RE GOING TOTALLY OPPOSITE DIRECTIONS. WOW.
There you go.
Most prevalent type of lie in social science today.
What just happened?
Will explain in the next post.
OpenAI Chat Log
So, then we paste in the exact data from the study,
and then do the 2 types of simple analysis we’re wondering about here:
(1) Regular linear regression — I.e. what today’s social “scientists” use to answer questions of the form:
“The smarter you are, the more likely you agree/disagree with X.”
(2) Midwit-aware linear regression — A slightly modified linear regression you’d need to use, were life full of midwit curves, to answer questions of the form:
“Once you cross a sufficiently smart threshold, where the getting smarter won’t make you flip your answer again anymore — then the more you continue toward unboundedly-high smartness, the more you agree/disagree with X.”
Basically, in both cases, we’re looking for:
“Smarter people believe X”.
So then we have ChatGPT run this analysis on the real data from the study
…and…
WTF IS THIS.
LOOK HOW OPPOSITE THOSE LINES ARE.
EVEN WORSE THAN I IMAGINED.
THEY’RE GOING TOTALLY OPPOSITE DIRECTIONS. WOW.
There you go.
Most prevalent type of lie in social science today.
What just happened?
Will explain in the next post.
OpenAI Chat Log
So, what happened here?
Why does is usual regression analysis that social scientists use (red line) — saying:
SMARTER -> LESS HESITANT
When clearly, with this midwit curve, the smartest people believe the complete opposite, and the more extreme you go into smartness, once you cross the midwit peak, the more that, clearly:
SMARTER -> MORE HESITANT?
Answer in next post.
Why does is usual regression analysis that social scientists use (red line) — saying:
SMARTER -> LESS HESITANT
When clearly, with this midwit curve, the smartest people believe the complete opposite, and the more extreme you go into smartness, once you cross the midwit peak, the more that, clearly:
SMARTER -> MORE HESITANT?
Answer in next post.
Here’s your answer
— Look at the red and yellow circles.
There are FAR MORE dumb people (red circle)
And FAR FEWER smart people (yellow circle)
So then, since the lying linear regression that social scientists love to use is WEIGHTED by the number of people in each bin — that makes the reversal in hesitancy on the right side practically dissapear, as far as the linear regression is concerned.
I.e. because there are always far fewer sufficiently smart people to fully understand why the midwit belief isn’t true, compared to the huge number of people on the rest of the curve — the high-iq portion essentialy completely dissapears, when using the type of naive linear regression analysis that social scientists always use.
=
Nearly everything controversial is midwit curves.
Nearly ever “smarter people believe X” claim made by social scientists is a lie,
because the method they use simply erases the upper-end of the midwit curve, simply because there’s so much fewer of those (yellow circle), than there is of everyone else.
They’re refusing to use midwit-compatible analysis methods.
Biggest lie in social science out there now.
It’s a lie that’s everywhere, because surprisingly, midwit curves are everywhere (Have posted about this in the past.)
— Look at the red and yellow circles.
There are FAR MORE dumb people (red circle)
And FAR FEWER smart people (yellow circle)
So then, since the lying linear regression that social scientists love to use is WEIGHTED by the number of people in each bin — that makes the reversal in hesitancy on the right side practically dissapear, as far as the linear regression is concerned.
I.e. because there are always far fewer sufficiently smart people to fully understand why the midwit belief isn’t true, compared to the huge number of people on the rest of the curve — the high-iq portion essentialy completely dissapears, when using the type of naive linear regression analysis that social scientists always use.
=
Nearly everything controversial is midwit curves.
Nearly ever “smarter people believe X” claim made by social scientists is a lie,
because the method they use simply erases the upper-end of the midwit curve, simply because there’s so much fewer of those (yellow circle), than there is of everyone else.
They’re refusing to use midwit-compatible analysis methods.
Biggest lie in social science out there now.
It’s a lie that’s everywhere, because surprisingly, midwit curves are everywhere (Have posted about this in the past.)
Forwarded from Chat GPT
Midwid Curve Confirmed, Yet Again!
The Inverse Scaling Prize identified eleven inverse scaling tasks, where worse performance was observed as a function of scale, evaluated on models of up to 280B parameters and up to 500 zettaFLOPs of training compute.
This paper takes a closer look at these inverse scaling tasks. We evaluate models of up to 540B parameters, trained on five times more compute than those evaluated in the Inverse Scaling Prize. With this increased range of model sizes and training compute, only four out of the eleven tasks remain inverse scaling. Six out of the eleven tasks exhibit what we call “U-shaped scaling”—performance decreases up to a certain model size, and then increases again up to the largest model evaluated.
Paper: Inverse scaling can become U-shaped
The Inverse Scaling Prize identified eleven inverse scaling tasks, where worse performance was observed as a function of scale, evaluated on models of up to 280B parameters and up to 500 zettaFLOPs of training compute.
This paper takes a closer look at these inverse scaling tasks. We evaluate models of up to 540B parameters, trained on five times more compute than those evaluated in the Inverse Scaling Prize. With this increased range of model sizes and training compute, only four out of the eleven tasks remain inverse scaling. Six out of the eleven tasks exhibit what we call “U-shaped scaling”—performance decreases up to a certain model size, and then increases again up to the largest model evaluated.
Paper: Inverse scaling can become U-shaped