6 ms·

There are no such thing as 210 IQ. It correlates to 7.3 sigma, meanwhile 7-sigma event has a probability of approximately 1 in 390 billion. We only have 8 bill

by Aperocky 11mo ago

There are no such thing as 210 IQ. It correlates to 7.3 sigma, meanwhile 7-sigma event has a probability of approximately 1 in 390 billion. We only have 8 billion humans on Earth. These absurd claims about IQ is almost evidence that the claimant are nowhere close. For starters, any IQ tests are not going to be normalized to that range because it is impossible to normalize to that range as there are 0 realistic samples.

paulddraper 11mo ago
That assumes a normal distribution.
- Aperocky 11mo ago
  There is no assumption, that is the definition of IQ. Is human intelligence a normal distribution? Probably not. But IQ is extrapolated as such, and probably useless anyway. Which make absurd claims like this even more laughable.
  - zahlman 11mo ago
    IQ is offset and scaled with the goal of producing that normal distribution, but that process is informed by the available data. Outliers don't get forced onto the curve, because there definitionally isn't enough data to figure out how outlying they are as a percentile.
    - Aperocky 11mo ago
      Exactly.
- airstrike 11mo ago
  I thought IQ was specifically designed to be normally distributed?
  - paulddraper 11mo ago
    That’s true. Also, this is where theory meets reality. Defining something as “the average” or “one std deviation “ is strange when the population is unmeasured and changing.
volkk 11mo ago
> It correlates to 7.3 sigma, meanwhile 7-sigma event has a probability of approximately 1 in 390 billion. We only have 8 billion humans on Earth. why can't you use historical population? like, the total amount of humans that ever existed? rough google shows around 100billion. seems legit that in the history of humanity, we could pop out someone so intelligent? But I do agree that IQ is probably a decent signal but entirely meaningless as sole measurement.
- JohnFen 11mo ago
  The other problem with IQ is that it's not a fixed scale, so you can't really compare IQ scores across time. An IQ of 100 is average by definition. Even if the average "intelligence" (or whatever IQ measures, because it doesn't seem to be intelligence as people think of intelligence) rises or falls over time, that average will always be a 100 IQ.
  - volkk 11mo ago
    why couldn't you? i've always imagined IQ as the raw potential? or am I misunderstanding
    - JohnFen 11mo ago
      Let's say, for the sake of argument, that people in the '60s were twice as smart as people now. The average IQ of the people then would be 100, and the average IQ of people now would also be 100, even though there was a huge difference in intelligence. This is because 100 is defined as being the average rather than being an absolute measure.
      - volkk 11mo ago
        Ah. okay yeah that makes sense. I didn't realize it's a relative measurement. I'm surprised it's not more robust. Something like using historical results to compare against, and updating tests in a very standardized way where the math/logic is always fairly similar, but the fact checking/knowledge that requires understanding of current world might be different data wise, but tests similar attributes or qualities.
airstrike 11mo ago
We only have 8 billion people on Earth _today_. Over 100 billion have lived through history, they say. Also just because it's statistically unlikely doesn't mean it can't happen. All else being equal, the very first human is just as likely to have had 210 IQ as the one born this morning.
- alternatex 11mo ago
  How can person have 210 IQ in isolation? The whole points system is based on standard deviation. What is this first person's IQ deviating from?
mr_mitm 11mo ago
It's more likely that someone screwed up on the test, or that he cheated somehow. Maybe he's articulate and good at solving logic puzzles, but his Wikipedia article clearly shows that the guy has a screw loose.
- Aperocky 11mo ago
  You can solve every one of the logic puzzles and it should not give you any score that high, unless it's a specifically designed bogus test to make certain person look and feel good. It's like saying your regular thermometer returned a reading of 1000C, sure buddy.
marginalia_nu 11mo ago
That's not really how statistics works. It is not impossible to roll two sixes on a single roll of two dies because it is more likely you won't.
- Aperocky 11mo ago
  It's not impossible, but improbable.
- nonethewiser 11mo ago
  That's right. If the event has a 1 in 390B chance, and there are 8B tries, you would expect to observe the event 2% of the time you conduct the 8B tries. So if you did 8B tries 50 times you would expect it to happen once. And its something higher than 8B - current population isn't the correct sample size.
  - Aperocky 11mo ago
    There's a lot of things that are theoretically possible, but to realistically consider them based on the known likelihood is something I'm not entertaining here.
  - mlyle 11mo ago
    That's assuming that the distribution is purely gaussian and nothing weird happens at the tails. I agree 276 is unlikely (and how would you even test/norm such a thing?)
- jcranmer 11mo ago
  IQ isn't some intrinsic property that we can measure that has a probability distribution. It's something we construct to have a specific probability distribution. Essentially, you have >145 IQ not because 0.1% people happen to have that high of an IQ but because we define 145 IQ to be the cutoff between the smartest 0.1% of the population and the dumbest 99.9%. As a consequence, an IQ above ~3 standard deviations is basically unmeasurable with more tests, and by 6 standard deviations I'm not sure is even possible to theoretically construct since I don't think what it's trying to measure can be meaningfully defined to discriminate the order of just 8 people on the entire planet.
Viliam1234 11mo ago
Everyone who actually studied psychometrics has to upvote this. (Which is why it stays at the bottom of the HN comment thread.)

paulddraper 11mo ago

That assumes a normal distribution.

Aperocky 11mo ago
There is no assumption, that is the definition of IQ. Is human intelligence a normal distribution? Probably not. But IQ is extrapolated as such, and probably useless anyway. Which make absurd claims like this even more laughable.
- zahlman 11mo ago
  IQ is offset and scaled with the goal of producing that normal distribution, but that process is informed by the available data. Outliers don't get forced onto the curve, because there definitionally isn't enough data to figure out how outlying they are as a percentile.
  - Aperocky 11mo ago
    Exactly.
airstrike 11mo ago
I thought IQ was specifically designed to be normally distributed?
- paulddraper 11mo ago
  That’s true. Also, this is where theory meets reality. Defining something as “the average” or “one std deviation “ is strange when the population is unmeasured and changing.

volkk 11mo ago

> It correlates to 7.3 sigma, meanwhile 7-sigma event has a probability of approximately 1 in 390 billion. We only have 8 billion humans on Earth. why can't you use historical population? like, the total amount of humans that ever existed? rough google shows around 100billion. seems legit that in the history of humanity, we could pop out someone so intelligent? But I do agree that IQ is probably a decent signal but entirely meaningless as sole measurement.

JohnFen 11mo ago
The other problem with IQ is that it's not a fixed scale, so you can't really compare IQ scores across time. An IQ of 100 is average by definition. Even if the average "intelligence" (or whatever IQ measures, because it doesn't seem to be intelligence as people think of intelligence) rises or falls over time, that average will always be a 100 IQ.
- volkk 11mo ago
  why couldn't you? i've always imagined IQ as the raw potential? or am I misunderstanding
  - JohnFen 11mo ago
    Let's say, for the sake of argument, that people in the '60s were twice as smart as people now. The average IQ of the people then would be 100, and the average IQ of people now would also be 100, even though there was a huge difference in intelligence. This is because 100 is defined as being the average rather than being an absolute measure.
    - volkk 11mo ago
      Ah. okay yeah that makes sense. I didn't realize it's a relative measurement. I'm surprised it's not more robust. Something like using historical results to compare against, and updating tests in a very standardized way where the math/logic is always fairly similar, but the fact checking/knowledge that requires understanding of current world might be different data wise, but tests similar attributes or qualities.

airstrike 11mo ago

We only have 8 billion people on Earth _today_. Over 100 billion have lived through history, they say. Also just because it's statistically unlikely doesn't mean it can't happen. All else being equal, the very first human is just as likely to have had 210 IQ as the one born this morning.

alternatex 11mo ago
How can person have 210 IQ in isolation? The whole points system is based on standard deviation. What is this first person's IQ deviating from?

mr_mitm 11mo ago

It's more likely that someone screwed up on the test, or that he cheated somehow. Maybe he's articulate and good at solving logic puzzles, but his Wikipedia article clearly shows that the guy has a screw loose.

Aperocky 11mo ago
You can solve every one of the logic puzzles and it should not give you any score that high, unless it's a specifically designed bogus test to make certain person look and feel good. It's like saying your regular thermometer returned a reading of 1000C, sure buddy.

marginalia_nu 11mo ago

That's not really how statistics works. It is not impossible to roll two sixes on a single roll of two dies because it is more likely you won't.

Aperocky 11mo ago
It's not impossible, but improbable.
nonethewiser 11mo ago
That's right. If the event has a 1 in 390B chance, and there are 8B tries, you would expect to observe the event 2% of the time you conduct the 8B tries. So if you did 8B tries 50 times you would expect it to happen once. And its something higher than 8B - current population isn't the correct sample size.
- Aperocky 11mo ago
  There's a lot of things that are theoretically possible, but to realistically consider them based on the known likelihood is something I'm not entertaining here.
- mlyle 11mo ago
  That's assuming that the distribution is purely gaussian and nothing weird happens at the tails. I agree 276 is unlikely (and how would you even test/norm such a thing?)
jcranmer 11mo ago
IQ isn't some intrinsic property that we can measure that has a probability distribution. It's something we construct to have a specific probability distribution. Essentially, you have >145 IQ not because 0.1% people happen to have that high of an IQ but because we define 145 IQ to be the cutoff between the smartest 0.1% of the population and the dumbest 99.9%. As a consequence, an IQ above ~3 standard deviations is basically unmeasurable with more tests, and by 6 standard deviations I'm not sure is even possible to theoretically construct since I don't think what it's trying to measure can be meaningfully defined to discriminate the order of just 8 people on the entire planet.

Viliam1234 11mo ago

Everyone who actually studied psychometrics has to upvote this. (Which is why it stays at the bottom of the HN comment thread.)